DNA-directed assembly of multivalent lipid nanoparticles for targeted T cell gene delivery

Abstract

Lipid nanoparticles (LNPs) are a powerful emerging tool for in vivo T cell engineering to treat diseases ranging from B cell lymphomas and other cancers to autoimmune diseases. Key challenges in designing these therapeutics include achieving both precise cell targeting and efficient mRNA translation. While single-targeted LNPs have been extensively studied, bispecific LNPs have only been briefly explored. Multiple receptor engagement offers the opportunity for enhanced mRNA delivery, expression, and T cell targeting. Here, a DNA-tethering method was developed to enable rapid modification of lipid nanoparticles with commercial antibodies. Using this strategy, we evaluated a variety of bispecific LNPs for targeted mRNA delivery to T cells both in vitro and in vivo. We identify bispecific formulations that improve targeting and subsequent transfection of T cells in vitro and in vivo relative to monotargeted LNPs. Additionally, we find that targeting molecules can alter LNP biodistribution to the spleen and liver. This fast and efficient approach to assembling antibody-targeted LNPs enables high-throughput screening of diverse antibody combinations for improved specificity and efficiency of in vivo gene delivery.

1. Introduction

Lipid nanoparticles (LNPs) have emerged as a potent and tunable platform for T cell engineering with vast potential to transform treatments of cancers, T cell infections, and autoimmune diseases [1-4]. For the treatment of B-cell lymphomas, chimeric antigen receptor (CAR) T therapies have been highly effective through laborious and expensive process of engineering T cells ex vivo [5,6]. There has been a shift towards in vivo generation of CAR T cells, which could be performed with off-the-shelf therapeutics that will streamline CAR T cell production and increase treatment accessibility. In vivo CAR’s have reached clinical trials for treating B-cell lymphomas, as well as autoimmune diseases, multiple myeloma, epithelial tumors, and liver cancer [7-10].

Key challenges for in vivo T cell gene therapies lie in the development of delivery vehicles capable of both highly efficient and on target T cell transfection [3,5,11,12]. LNP delivery approaches are currently limited to localized administration and hepatic cells, which have a natural propensity for particle uptake. To overcome this limitation, two main strategies have been pursued to redirect LNP biodistribution: [1] optimizing the structure of ionizable lipids to achieve tropism for non-hepatic cells and organs, and [2] surface-modifying LNPs with receptor-specific antibodies for targeted delivery [1,13-15]. The latter approach has been explored using various single receptor targets, including CD3 [16,17], CD4 [18,19], CD5 [16,20], CD7 [16,21], and CD8 [22,23]. These formulations generally result in low transfection efficiency, which may be attributed to poor or partial T cell activation, —a factor shown to be key for the expression of LNP-delivered mRNA [24,25]. In the body, T cell activation is controlled by interactions with antigen presenting cells (APCs). Activation is a multivalent phenomenon, achieved through providing a primary activation signal, usually through engagement with the T cell receptor (TCR) complex in combination with a costimulatory signal [26]. Most often, a combination of antibodies against CD3 and CD28 are used to activate T cells ex vivo to mimic the stimulation provided by APCs [27]. In previous work, bispecific LNPs modified with a combination αCD3 and αCD28 antibodies have been explored for ex vivo T cell engineering and showed an increase transfection in comparison to LNPs modified with αCD3 alone [24]. Some bispecific formulations such as αCD3/αCD7-LNPs have also seen success for in vivo CAR generation to treat B cell leukemia [25]. Further screening of bispecific LNPs both ex vivo and in vitro could identify new T cell surface targets that increase efficacy of mRNA delivery. An understanding of how co-targeting use of immunomodulatory signals impacts T cell transfection is key for therapeutic optimization and requires methods to rapidly assemble and screen many combinations of targeted LNPs (tLNPs).

Several technical barriers in antibody conjugation have limited progress in development of tLNPs. Traditional methods for conjugating antibodies to the surface of LNPs such as amide formation and maleimide chemistry are often time-consuming, result in low efficiency, and lead to incorrect antibody orientation which is not conducive for screening large combinations of targeting molecules [28]. While click chemistry offers a rapid method to attach multiple antibodies, it lacks sufficient orthogonal attachment chemistry to control the ratios or spatial arrangement of attached proteins [29,30]. Other innovative approaches have been developed to address these limitations. For example, Peer et al. employed ASSET, a membrane-anchored lipoprotein that specifically binds the Fc domain of antibodies, to enable rapid and oriented surface modification [31]. Previous work in our laboratory has also utilized self-labeling enzyme fusion proteins to efficiently functionalize particles [32]. Building on these strategies, oligonucleotide- based assembly has emerged as a robust, rapid, and highly precise method for tethering antibodies to nanoparticles [33,34]. By leveraging the highly specific base-pairing interactions between complementary nucleic acid sequences, we can design and control specific material and biomolecule interactions [35-40].

Here, we developed a workflow for the rapid modification of LNPs using DNA-tethering of commercially available antibodies for screening combinations of antibodies for enhanced transfection efficiency in T cells. T cells were chosen as a model due to their therapeutic relevance, resistance to transfection, and the well-characterized expression of specific surface markers across their subsets. We demonstrate the use of bispecific LNPs for both in vitro delivery to Jurkat cells and primary human T cells and in vivo screening of top-performing formulations. By leveraging this DNA-anchoring method for antibody conjugation to LNPs, we can rapidly identify antibody combinations capable of efficiently targeting and transfecting T cells, paving the way for the development of highly effective in vivo T cell engineering for multiple pathologies.

2. Materials and methods

2.1. Materials

The following chemicals were used as received: Acetic acid, glacial, 99 + % (AA33252AK, Thermo Scientific Chemicals), chloroform, stabilized with amylene for HPLC (C297–4, Fisher Chemical), ethanol, absolute (200 proof), molecular biology grade (BP28184, Fisher Bio-Reagents), methanol (A452SK-4, Fisher Chemical), and EDTA (AM9260G, Invitrogen). 1,2-distearoyl-sn-glycero-3-phosphoethanol-amine-N-[azido(polyethylene glycol)-2000] (ammonium salt) (DSPE-PEG(2000) Azide, 880228), cholesterol (ovine, 700000), 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)-2000] (ammonium salt) (DSPE-PEG2000, 880120), 1.2-distearoyl-sn- glycero-3-phosphocholine (DSPC, 850365), 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[amino(polyethylene glycol)-2000]-N-(Cyanine 7) (DSPE PEG(2000)-N-Cy7, 810892), and 1,2-dioleoyl-sn- glycero-3-phosphoethanolamine-N-(lissamine rhodamine B sulfonyl) (ammonium salt) (18,1 Liss Rhod PE, 810150) were purchased from Avanti Polar Lipids. D-Lin-MC3-DMA (MC3, BP-25497) was purchased from BroadPharm. SYBR gold nucleic acid gel stain (S11494), Quant-iT OliGreen ssDNA Reagent and Kit (O11492), and Quant-it RiboGreen RNA Assay Kit (R11490) were purchased from Invitrogen. 2-mercaptoe- thanol (1610710), QC colloidal Coomassie (1610803), precision plus protein dual color standards (1610374), and 4–20% polyacrylamide gradient gel (4561096) were purchased from Bio-Rad. Dulbecco’s phosphate-buffered saline, no calcium, no magnesium (dPBS, 14190), RPMI 1640 Medium (11875093), penicillin streptomycin (pennstrep, 15–140-122) and fetal bovine serum, heat inactivated (A5669801) were purchased from Gibco. 1× red blood cell lysis buffer (00–4333-57) was purchased from eBioscience. Sepharose 4B (4B200) and Sepharose CL-2B (CL2B300) were purchased from Sigma-Aldrich. OYo-Link custom oligo reagent (5′ TTA ATC ACC CTC GCG CAC TAC 3′, AT1002) for modification for general antibody modification and for mouse IgG1 modification was purchased from AlphaThera. 5′ -/5ATTO633N/TTA ATC ACC CTC GCG CAC TAC 3′ and 5′ /5DBCON/ CTA GTG CGC GAG GGT GA 3′ were purchased from IDT. For surface modification of LNPs for in vitro studies, Alexa Fluor 647 anti-human CD28 antibody (302953), Ultra-Leaf purified anti-human CD3 antibody (300438), purified anti-human TCR α/β antibody (306702), purified anti-human CD2 antibody (300202), Ultra-LEAF purified anti-human CD4 antibody (300569), purified anti-human CD5 antibody (300602), purified anti-human CD8 antibody (344702), Ultra-LEAF purified anti-human CD28 antibody (302934), and purified moused IgG1, κ isotype control antibody (400102) were purchased from Biolegend. Purified mouse anti-human CD6 (555356) and purified mouse anti-human CD7 (555359) were purchased from BD Biosciences. For surface modification of LNPs for in vivo studies, CD3e monoclonal antibody (16–0031-82), TCR beta monoclonal antibody (16–5961-82), CD2 recombinant rabbit monoclonal antibody (18056481), CD4 monoclonal antibody (16–0041-82), and rat IgG2b κ isotype control (16–4888-81) were purchased from Invitrogen. ONE-Glo + Tox Luciferase Reporter and Cell Viability Assay (E7120) and VivoGlo Luciferin (P1043) were purchased from Promega. mCherry mRNA (L-7203) and luciferase mRNA (L-7602) were purchased from TriLink. Saline sodium chloride 0.9% (04888010) was purchased from Hospira. For flow cytometry studies on human cells, Zombie UV Fixable Viability Kit (423107), Human TruStain Fc receptor blocking solution (422302), APC-Cyanine7 anti-human CD3 antibody (344818), CD4 Pac Blue 317,429 Pacific Blue anti-human CD4 antibody (317429), CD8 FITC anti-human CD8a antibody (300906), CD14 APC anti-human CD14 antibody (325607), and CD19 PE/Cyanine7 anti-human CD19 antibody (302216) were purchased from Biolegend. For flow cytometry studies on mouse cells, TruStain FcX PLUS (anti-mouse CD16/32, 156603), Spark Red 718 anti-mouse CD3 antibody (100282), Alexa Fluor 647 anti-mouse CD4 antibody (100426), Alexa Fluor 488 anti-mouse CD8a antibody (100726), Brilliant Violet 421 anti-mouse CD19 antibody (115537), and PE anti-mouse/human CD11b antibody (101207) were purchased from Biolegend. Toxicity assays included Mouse IL-6 Uncoated ELISA Kit (88–7064-88, Invitrogen), Mouse TNF alpha ELISA Kit (BMS607–3), and Bilirubin Assay Kit (MAK126, Sigma- Aldrich).

2.2. Antibody modification and characterization

Full length Immunoglobulin G (IgG) antibodies were conjugated to single stranded DNA using light activated site-specific conjugation [41]. oYo-Link mIgG1 was used for all IgG1 antibodies with the mouse host species. All other antibodies were modified with oYo-Link. 100μg of oYo-Link custom oligo reagent (5′ TTA ATC ACC CTC GCG CAC TAC 3′) was rehydrated in 100 uL of ultrapure water. 1 uL of oYo-Link reagent was mixed with 1 μg of antibody and then placed under ~365 nm UV light (AT8001-D, AlphaThera) for 2 h.

A reducing SDS-PAGE gel was used to analyze antibody modification. Samples were prepared at 0.5 uM antibody in with 10% volume of 2-mercaptoethanol. The samples were heated for 10 min at 90 °C. 10 uL of each sample was loaded into a 4–20% polyacrylamide gradient gel. 3 uL of the protein ladder was loaded into well 1. The gels ran for ~1 h at 120 V. The gel was incubated in 1× SYBR Gold for 10 min before imaging using AzureRed fluorescent imaging (Azure 200, Azure Biosystems). The gel was then fixed in a solution of 50% v/v methanol and 1% v/v acetic acid for 15 min before incubating in colloidal Coomassie for 2 h. Gels were imaged using Coomassie blue settings (Azure Biosystems).

2.3. DNA-lipid synthesis

As previously described by Banga et al. [39], DNA conjugated lipids were synthesized using copper-free click chemistry. Briefly, the DBCO-terminated oligonucleotide (5′ /5DBCON/ CTA GTG CGC GAG GGT GA 3′) was dissolved in 100 uL ultrapure distilled water at a concentration of 100 uM. In a separate glass vial, 100 nmol (28 uL at 10 mg/ mL) of DSPE-PEG(2000) Azide in chloroform was dried under inert nitrogen gas and then dissolved in 100 uL ethanol. The aqueous oligonucleotide solution was then added to the lipid solution and the mixture was left to shake on the benchtop at room temperature. The next day the contents of the vial were dried under inert nitrogen gas. The dried material was resuspended in 300 uL ultrapure distilled water. Excess lipid was purified using chloroform extraction. Briefly, 300 uL of chloroform was added to the solution and mixed through gentle inversion before spinning for 5 min at 16,000 g and removing the top aqueous layer. This chloroform wash was repeated 3 times. Concentration of DNA-lipid was determined using Quant-iT OliGreen ssDNA Reagent and Kit according to the manufacturer’s instructions. The molecular weight of the DNA-lipid was confirmed using mass spectrometry at the IMSERC Mass Spectrometry Facility.

2.4. LNP formulation and characterization

LNPs were prepared at the molar ratio of 50:13:35:2 of MC3:DSPC: cholesterol:DSPE-PEG2000. For fluorescent LNPs, the 2 mol% DSPE- PEG2000 was replaced with 1.7 mol% DSPE-PEG2000 and 0.3 mol% DSPE PEG(2000)-N-Cy7 or 18:1 Liss Rhod PE. Lipids were dissolved in ethanol to form the ethanol phase. 30 uL mRNA at a concentration of 1 mg/mL or 30 uL PBS for empty LNPs, 22.5 uL citrate buffer at a concentration of 100 mM, and 172.5 uL ultrapure water were combined in a glass vial to form the aqueous phase using a stir bar and by manual pipette mixing. Briefly, the aqueous phase was mixed using a stir bar and stir plate (Fisher Scientific) at 700 rpm while the ethanol phase was added dropwise to the aqueous phase. LNPs were dialyzed overnight at 4 °C.

Encapsulation efficiency was measured for LNPs using Quant-it RiboGreen RNA Assay Kit according to the manufacturer’s instructions. LNP size, polydispersity, and zeta potential were determined using dynamic light scattering (DLS, Malvern Zetasizer Ultra). 30 uM LNP were read for each DLS experiment, and each replicate is the average of 3 reads of a single sample. To measure zeta potential samples were diluted 1:1000 and suspended in 0.1xPBS buffer, and each replicate is the average of 3 reads of a single sample.

2.5. Ab-LNP conjugation and characterization

LNPs were modified through incubation of oYo-Link modified antibody at a ratio of 0.02 uM/1 mM LNP lipid, DNA-lipid at a ratio of 1.8 uM/1 mM LNP lipid, and LNPs at 37 °C for 1 h. For bispecific LNPs, modified antibodies were added at a 1:1 ratio with the ratio of 1.8 uM/1 mM LNP lipid remaining the same. For transfection of human cells in vitro (Jukat cells and PBMCs), LNPs were modified with anti-human CD3, anti-human TCR, anti-human CD2, anti-human CD4, anti-human CD5, anti-human CD6, anti-human CD7, anti-human CD8, anti-human CD28, and IgG1 κ isotype control. For transfection of mouse cells in vivo, CD3e monoclonal antibody, TCR beta monoclonal antibody, CD2 recombinant rabbit monoclonal antibody, CD4 monoclonal antibody, and rat IgG2b κ isotype control were used.

Size exclusion chromatography was used to determine DNA-lipid insertion into and antibody conjugation to LNPs. Following LNP modification with DNA-lipid, a complementary fluorescent ssDNA (5′-/5ATTO633N/TTA ATC ACC CTC GCG CAC TAC 3′) was incubated in excess with DNA-LNPs at 37 °C for 1 h. DNA-LNP samples were loaded into a column packed with Sepharose 4B and fractions were collected in a 96 well plate using a Gilson FC204 Fraction Collector. Plates were then read on a Molecular Devices Spectra Max iD3 plate reader for ATTO 633 (ex 620 nm/em 660 nm) and Liss Rhod (ex 550 nm/em 590 nm). For Ab-LNPs, an Alexa Fluor 647 anti-human CD28 antibody was conjugated to the surface. Ab-LNPs were loaded into a column packed with Sepharose CL-2B. Fractions were collected as previously described. Plates were read for CD28 antibody (ex 630/ em 670) and Cy7 (ex 745 nm/em 785 nm). To determine the concentration of DNA or antibody conjugated to the LNP, the measured fluorescence in LNP (void volume) fractions was compared to a standard curve for each fluorophore. LNP concentrations were determined using nanoparticle tracking analysis. Samples with a lipid concentration of 0.15 mM were analyzed using NanoSight300 (Malvern Instruments).

Agarose gel electrophoresis was performed to assess the encapsulation efficiency of DNA-modified LNPs. Samples were run on a 2% agarose gel containing SYBR Gold nucleic acid stain. For each lane, 10 μL of LNPs (50,000 ng mRNA/mL) were loaded. For lysed samples, 1 μL of 10% Triton X-100 was added prior to loading to disrupt the particles and release encapsulated mRNA. Gels were imaged using the Azure 200 imaging system (Azure Biosystems). Band intensity was quantified using Fiji (ImageJ).

2.6. Cell culture

Jurkat cells were purchased from ATCC (TIB-152). PBMC’s were generously gifted from the Choi Lab (Northwestern Medicine) or purchased from ATCC (PCS-800-011). Cells were cultured in RPMI 1640 Medium supplemented with 10% FBS and 1% pennstrep. NFAT–Jurkat cells were transfected with an NFAT-reporter gene as previously described [32].

2.7. In vitro Jurkat cell luciferase and toxicity assay

Jukat cells were plated in triplicate at 100,000 cells per 50 uL supplemented RPMI. 50 uL LNPs suspended in PBS (containing 750 ng luciferase mRNA) were added to each well. After 24 h incubation at 37 °C, ONE-Glo + Tox Luciferase Reporter and Cell Viability Assay was used to determine cell luminescence and LNP toxicity according to the manufacturer’s instructions. Briefly, 20 uL 5× CellTiter-Fluor Reagent to each well and mix using orbital shaking for ~30 s. After 30-min incubation at 37 °C, fluorescence was measured on a Molecular Devices Spectra Max iD3 plate reader (ex 465 nm/em 505 nm). Next, 100 uL ONE-Glo Reagent was added to each well and was incubated for 3 min at room temperature before luminescence was measured on a Molecular Devices Spectra Max iD3 plate reader. For the 72-h assays, Jurkat cells were treated with LNPs for 24 h, then resuspended in supplemented RPMI and cultured for an additional 48 h.

2.8. In vitro Jurkat cell activation

NFAT–Jurkat cells maintained in supplemented RPMI were seeded at 1 × 10⁵ cells per well in 50 μL. LNPs or tLNPs diluted in 50 μL PBS were added at a dose of 750 ng mRNA per 1 × 10⁵ cells, resulting in a final well volume of 100 μL. After 24 h of incubation, cells were washed three times with flow buffer (dPBS supplemented with 1% FBS and 2 mM EDTA) and resuspended in flow buffer for analysis. eGFP expression was quantified using an LSR Fortessa 2 flow cytometer (BD Biosciences).

2.9. In vitro Jurkat cell binding and internalization

Jurkat cells maintained in supplemented RPMI were seeded at 1 × 10⁵ cells per well in 50 μL. Rhodamine-labeled LNPs or tLNPs lacking mRNA were diluted in 50 μL PBS and added to cells at a dose equivalent to 750 ng total RNA formulation per 1 × 10⁵ cells. After 3 h of incubation at 37 °C, cells designated for the internalization-only group were washed with PBS and treated with 0.25% trypsin–EDTA for 10 min to remove surface-bound particles. Cells in the binding + internalization group were not treated with trypsin–EDTA. All samples were subsequently washed with PBS and stained with Zombie UV Fixable Viability Kit according to the manufacturer’s instructions. Cells were washed three times with flow buffer (dPBS supplemented with 1% FBS and 2 mM EDTA) prior to analysis on an LSR Fortessa 2 flow cytometer (BD Biosciences).

2.10. In vitro PBMC transfection

PBMC’s were thawed in supplemented RPMI, rested for 24 h, and subsequently plated in a round bottom 96 well plate at 300,000 cells per 100 uL media. 100 uL LNPs suspended in PBS containing 2000 ng mCherry mRNA were added to each well. After 24 h incubation at 37 °C, cells were suspended in dPBS and stained with Zombie UV Fixable Viability Kit according to the manufacturer’s instructions. Cells were then suspended in 100 uL flow buffer and blocked using Human TruStain Fc receptor blocking solution. Cells were stained with APC-Cyanine7 anti-human CD3 antibody, Pacific Blue anti-human CD4 antibody, FITC anti-human CD8a antibody, APC anti-human CD14 antibody, and PE/Cyanine7 anti-human CD19 antibody for 30 min at 4 °C. Stained cells were analyzed on LSR Fortessa 2 Analyzer (BD Biosciences) and data sets were analyzed on FlowJo 10.10.0 (BD Biosciences). For the PBMC dose response, cells were plated in triplicate at 100,000 cells per 50 uL supplemented RPMI. 50 uL LNPs suspended in PBS were added to each well. After 24 h incubation at 37 °C, ONE-Glo + Tox Luciferase Reporter and Cell Viability Assay was used to determine cell luminescence and LNP toxicity according to the manufacturer’s instructions.

2.11. Animals

All animal studies were conducted according to the guidelines for the Care and Use of Laboratory Animals from the National Institutes of Health [42] and all animal work was performed under protocol IS00005877 approved by the Institutional Animal Care and Use Committee (IACUC). Male BALB/c mice ~10–12 weeks in age were purchased from Jackson Labs used in all animal studies. Mice were placed on a low-fluorescent diet several days prior to all experiments.

2.12. In vivo biodistribution using luciferase mRNA

LNPs containing a Cy7 dye and luciferase mRNA were given via tail vein injection at a dose of ~0.6 mg firefly luciferase (Fluc) mRNA/kg body weight. The volume of LNPs to inject was calibrated to the body weight of each mouse. In vivo images were taken using an in vivo imaging system (IVIS) Spectrum (PerkinElmer, Waltham, MA) for the fluorescence intensity of Cy7 (ex 745 nm/em 780 nm) and bioluminescence of Fluc at 0.5, 5, and 24 h. D-luciferin in saline was given at 75 mg/kg body weight via intraperitoneal injection 10 min prior to imaging. Blood was collected into EDTA-treated tubes (365,974, Fisher Scientific) using retroorbital bleeding at 0.5 and 5 h. 24 h following tail vein injection, mice were euthanized, and blood was collected using cardiac puncture. The left and right inguinal lymph nodes, spleen, liver, heart, kidneys, lungs, skin, and leg were imaged ex vivo for Cy7 fluorescence and bioluminescence of luciferase using the IVIS Spectrum (PerkinElmer) and the mass of each organ was recorded.

2.13. Analysis of spleen histology

Spleen histological images were analyzed using Fiji. Color deconvolution was performed using the H&E vector for hematoxylin and eosin (H&E)–stained sections and the H DAB vector for CD4- and CD8-stained sections. Three regions of interest (ROIs) were selected per image for quantitative analysis. For H&E-stained samples, background subtraction was performed using the rolling ball algorithm (radius = 100 μm) prior to thresholding. Quantification was conducted on the blue channel following thresholding at 110–255 to define white pulp regions. Red pulp area was calculated by subtracting the measured white pulp area from the total ROI area. For CD4- and CD8-stained sections, color channel 2 from the H DAB deconvolution was selected for analysis. Thresholding was applied on a per-image basis using the “Percentile Dark” method for CD4 (due to higher staining density) and the “Default Dark” method for CD8 (due to lower staining density and greater particle separation). CD4- and CD8-positive staining was quantified as percentage area of the total ROI.

2.14. In vivo delivery using mCherry mRNA

LNPs containing a Cy7 dye and mCherry mRNA were given via tail vein injection at a dose of ~0.6 mg Fluc mRNA/kg body weight. 6 h after LNP injection, blood was collected using heart puncture into EDTA- treated tubes. Red blood cells were lysed using 1× red blood cell lysis buffer. The spleen and left and right inguinal lymph nodes were collected and pressed through a 40-um cell strainer (08–771-1, FisherScientific) to create a single cell suspension. Cell suspensions were stained using Zombie UV Fixable Viability Kit according to the manufacturer’s instructions. Cells were then suspended in 100 uL flow buffer and blocked using TruStain FcX PLUS. Cells were stained with Spark Red 718 anti-mouse CD3 antibody, Alexa Fluor 647 anti-mouse CD4 antibody, Alexa Fluor 488 anti-mouse CD8a antibody, Brilliant Violet 421 anti-mouse CD19 antibody, and PE anti-mouse/human CD11b antibody. Samples with low T cell counts (<500) were omitted from analysis.

2.15. Cytokine and bilirubin assays

Whole blood was centrifuged to isolate serum. Interleukin-6 (IL-6) and tumor necrosis factor α (TNF-α) levels were quantified using ELISA kits according to the manufacturers’ instructions. Total bilirubin was measured using a colorimetric assay performed according to the manufacturer’s protocol.

2.16. Statistics

Statistical analyses were performed using Prism 10.4.0 software (Graph-Pad Software Inc.) Statistical tests are reported within each figure caption. P-values of <0.05 were considered significant and were adjusted for multiple comparisons. Data are expressed as mean ± SEM, unless otherwise specified.

3. Results and discussion

3.1. DNA-tethering enables rapid and efficient attachment to LNP surface

A DNA-tethering conjugation method was used to attach antibodies to the surface of LNPs (Fig. 1A). First, LNPs were synthesized using the ionizable cationic lipid Dlin-MC3-DMA, which was also used in the first FDA-approved LNP gene therapy [43]. Oligonucleotide-conjugated lipid (DNA-lipid) was synthesized using copper-free click chemistry [39] and confirmed using mass spectrometry (SI Appendix, Fig. S1). A complementary strand of ssDNA was attached to the antibodies of interest (DNA-Ab) using a photoactivated Protein G which enables site specific modification of nearly all IgG antibodies with high efficiency (AlphaThera) [41]. Successful DNA-Ab synthesis was confirmed using gel electrophoresis (SI Appendix, Fig. S2). The DNA-lipid was post-inserted into preformed LNPs and the insertion was assessed using a fluorescent complementary DNA (cDNA) followed by size exclusion chromatography (SEC) (Fig. 1B-D and SI Appendix, Fig. S3). We compared the use of a DNA-lipid strand containing a 15 atom tetraethylene glycol (TEG) spacer to a DNA-lipid without the spacer (SI Appendix, Table S1). We found a greater than 2-fold increase in the average amount of conjugated cDNA using the DNA-lipid without the TEG spacer (Fig. 1B). This difference in conjugation efficiency indicates that the TEG spacer either hinders the lipid post-insertion process or sterically blocks the binding of the cDNA strand. All further studies were completed without the TEG spacer.

Fig. 1.

Method for DNA-tethering of antibodies to LNPs and particle characterization. (a) Schematic of modification of LNPs through a 1-h incubation of LNP, DNA-lipid, and DNA-Ab at 37 °C. (b) DNA-LNPs were incubated with a fluorescent Atto647 cDNA strand and run on SEC. The nM cDNA per mM of LNP membrane were quantified. DNA-lipid was added at a concentration of 1 uM with and without a TEG spacer (error bars = standard error of mean) and DNA-lipid was added at a concentration of 5 uM. The modified particles were incubated for 30 min at 37 °C with 0.5 U/mL (+) and 10,000 U/mL (++) DNase I before SEC (error bars = standard error of mean). Two outliers were removed using ROUT using Q = 10%. (c) DNA-LNPs were incubated with a fluorescent AF647 DNA-Ab and run on SEC. The nM of antibody per mM of LNP membrane was quantified. This was also measured after incubation with 0.5 U/mL DNase I (error bars = standard error of mean). All samples incubated with DNase were performed in triplicate (n = 3), whereas all other experimental conditions were conducted with five replicates (n = 5). (d) Particle size was measured using DLS for LNPs before and after modification (n = 3). (f) The z-average, PDI, and zeta potential with SEM for the unconjugated LNPs, DNA-LNPs and tLNPs are shown (n = 3).

We next assessed the stability of our DNA-tethers in the presence of physiologically relevant deoxyribonuclease (DNase) concentrations. For this modification method to be useful in vivo, it would be necessary for the attachment strategy to remain stable. A DNase I concentration of 0.5 U/mL was used, which falls slightly above the average DNase activity in human plasma reported by Tamkovich et al. [44] of 0.356 ± 0.410 U/mL. As a positive control for DNase I activity a concentration of 10,000 U/mL was used to confirm DNase activity. It was found that at the concentration of 0.5 U/mL, the cDNA strands remained attached to the particle surface whereas at high DNase concentrations of 10,000 U/mL the cDNA was almost completely cleaved as assessed by SEC (Fig. 1B).

Following confirmation of DNA stability, we quantified the attachment of the DNA-Ab to the surface of our particle. We used 5 μM of DNA-lipid and 0.06 μM of fluorescent AF647 CD28 DNA-Ab to modify our LNPs. After 1 h, these synthesized particles were run on SEC and the average nM of antibody per mM of LNP lipid was found to be 10.7 ± 1.3 (Fig. 1C). This value did not change significantly after a 30 min incubation period with 0.5 U/mL DNase I, indicating that surface modified nanoparticle is stable at physiological DNase levels [33]. Using NanoSight300 (Malvern), it was determined across 4 LNP batches containing either mCherry or Fluc mRNA that the concentration was 6.33 × 10¹¹ ± 1.2 × 10¹¹ particles/mL (SI Appendix, Fig. S4 and Table S2). Using the concentration of antibody from SEC in combination with this concentration of particles, it can be estimated that there are approximately 33 ± 3 antibodies per LNP which is approximately 58% of the theoretical maximum amount of antibody added to solution.

The size, zeta potential, and encapsulation efficiency of the synthesized LNPs were also characterized to ensure proper LNP formation and stability. Using DLS the particle size of Fluc LNPs was estimated to be 156.1 ± 2.6 nm with a PDI of 0.14 ± 0.01 (Fig. 1D-E). The measured diameter increased to approximately 310.9 ± 5.2 nm with a PDI of 0.2 ± 0.01 for the DNA-lipid modified LNP indicating particle aggregation. The tLNP had a size of 182.2 ± 1.1 with a PDI of 0.18 ± 0.02, an increase of ~26 nm from the unmodified particle accounting for the addition of the antibody to the particle surface. Further confirmation of approximate LNP size and spherical shape was performed using CryoEM (SI Appendix, Fig. S5). The encapsulation efficiency of the Fluc LNPs was 79.53 ± 0.013% using the Quant-it RiboGreen RNA Assay Kit which aligns with other studies using LNPs prepared using manual mixing [45]. The DNA oligos used to modify the particles interfere with this assay. To determine the encapsulation after DNA-lipid post insertion and antibody attachment, we ran agarose gel electrophoresis and found no change in encapsulation after particle modification (SI Appendix, Fig. S6). The zeta potential of the unmodified LNP was found to be −12.7 ± 0.06 mV which aligns with other reports for Dlin-MC3-DMA LNPs (SI Appendix, Fig. S7) [46]. Upon addition of the DNA-lipid, there was a decrease in zeta potential to −20.1 ± 1.3 mV, as expected due to the negatively charged phosphodiester backbone of the DNA. After addition of the antibody the zeta potential was unchanged at −19.6 ± 1.0 mV [47]. LNPs were then used for in vitro and in vivo assessment of mRNA delivery to T cells. Antibody conjugation using our DNA-based therefore had minimal alterations to the LNP’s physical properties.

3.2. DNA-mediated antibody conjugation to LNPs improves transfection of Jurkat cells

For initial evaluation of LNP transfection in vitro, Jurkat cells were transfected using single-targeted and bispecific LNPs. All Jurkat cell experiments were performed with LNPs encapsulating mRNA encoding for Fluc. We tested the transfection levels and viability of Jurkat cells as a function of mRNA dose for five LNP variations to establish the functionality of the conjugation strategy (Fig. 2). The potency of αCD3-tLNPs has been well established in literature [16,17,24] so this targeting molecule was used as a model in initial validation studies and as a control throughout all subsequent experiments. Overall, Fluc expression increased with LNP dose (Fig. 2B). Notably, the addition of DNA to the surface of the particles and the conjugation of an isotype, nonbinding antibody caused a decrease in transfection relative to the unconjugated LNP. This may be attributed to the negative charge of the DNA on the LNP surface which can cause electrostatic repulsion between the LNP and the cell membrane, hindering delivery [48,49]. The addition of soluble αCD3 antibody did not cause a change in luminescence relative to the unconjugated LNP. In comparison, when the αCD3 was properly conjugated to LNP surface (αCD3-LNP) the highest luciferase expression occurred. The viability appeared to decrease with increased dosing for the isotype-LNP and αCD3-LNP groups at ~75% for the largest tested dose of 1000 ng/100,000 cells (Fig. 2C). A mRNA dose of 750 ng/100,000 cells was used for all subsequent Jurkat cell experiments. To understand the longevity of LNP toxicity and mRNA expression, cells were treated with LNP’s and cultured for a total of 72 h. After, the αCD3-LNP group again had the highest luciferase expression, and all groups maintained a viability of greater than 90% likely indicating proliferation of the T cells after LNP treatment (SI Appendix, Fig. S8).

Fig. 2.

mRNA dose response curve and testing of monospecific and bispecific DNA-tethered tLNPs. (a) Schematic of LNPs tested in the dose response. (b) Luciferase expression of Jurkat cells treated with luciferase-encoding mRNA using a range of mRNA doses after 24 h (n = 3 replicates, error bars = standard error of the mean). (c) Viability of Jurkat cells treated with luciferase-encoding mRNA using a range of mRNA doses after 24 h (n = 3 replicates, error bars = standard error of the mean). (d) Luciferase expression in Jurkat cells after treatment with tLNPs with and without the DNA-lipid attachment after 24 h with a dose of 750 ng/100,000 cells. (e) Viability of Jurkat cells after treatment with tLNPs with and without the DNA-lipid attachment after 24 h with a dose of 750 ng/100,000 cells (n = 3 replicates, error bars = standard error of the mean). Statistical analysis included 2-way ANOVA with Dunnett’s multiple comparisons test, *p-value <0.05 as compared to αCD3-LNP. (f) Heat map of Jurkat luciferase expression after 24 h treatment with bispecific LNPs at a dose of 750 ng/100,000 cells. (g) Jurkat NFAT eGFP reporter cell line expresses eGFP when incubated with tLNPs and soluble antibodies (n = 3 replicates, error bars = standard error of the mean).

Next, antibodies targeting individual common T cell surface receptors were screened for their capacity to increase LNP-mediated transfection of Jurkat cells. Each formulation was tested with and without the DNA-lipid to confirm that the conjugation strategy was necessary for high transfection rather than mediated by soluble antibody in solution. In each case, the presence of the DNA-lipid linker on an LNP improved Jurkat cell transfection (Fig. 2D). Interestingly, the isotype-LNP showed the opposite trend where the addition of the linker decreased overall transfection following a similar trend as seen in the dose response where addition of the DNA-lipid without a T cell targeting antibody decreases overall transfection (SI Appendix, Fig. S9). The highest T cell transfection was seen with αTCR, αCD3, αCD5 and αCD28-LNPs. Notably, both αTCR and αCD3 engage with proteins which are part of the TCR complex while aCD28 is a commonly used costimulatory molecule [24]. Jurkat cell viability remained at ~75% for αTCR and αCD3-LNPs but was significantly higher for the other antibodies tested (Fig. 2E). It should be noted that the choice of antibody clone affected luciferase expression, likely reflecting the importance of affinity and/or epitope specificity of selected antibodies in determining transfection efficacy (SI Appendix, Fig. S10).

We next determined if the combinatorial use of these antibodies could improve the transfection of the Jurkat cells. Using a 1:1 ratio of antibodies, we screened 36 different combinations of the antibodies on the surface of LNPs. We found that combinations that included either αTCR or αCD3 transfected Jurkat cells with Fluc more efficiently than other combinations. Notably αTCR/αCD2, αCD3/αCD2, αCD3/αTCR, and αCD3/αCD28-LNP had the highest luciferase expression of the 36 combinations screened (Fig. 2F and SI Appendix, Fig. S11). For future studies, αCD3/αTCR, αCD3/αCD2, and αCD3/αCD28-LNPs were chosen for screening due to their high transfection in Jurkat cells. In addition, αCD3/αCD4-LNPs were chosen because of the potential to selectively target CD4⁺ T cells which are relevant to CAR T cell and autoimmune cell therapeutics and αCD3/αCD5-LNPs were chosen due to the potential of αCD5 engagement to improve alter T cell function [18,20].

We next sought to determine how these selected LNP formulations influence T cell activation. tLNPs or soluble antibodies were incubated with a Nuclear Factor of Activated T Cells (NFAT) reporter Jurkat cell line [32,50]. This reporter line contains an enhanced green fluorescent protein (eGFP) gene downstream of an NFAT response element, such that T cell activation leads to NFAT translocation and eGFP expression. eGFP expression in treated cells was quantified by flow cytometry. We observed that control groups (PBS, LNP, and isotype-LNP) did not induce Jurkat cell activation, whereas groups containing functional T cell–binding antibodies increased eGFP expression (Fig. 2G and SI Appendix, Fig. S12). For all targeted groups, antibody presentation on the LNP surface resulted in greater activation than the corresponding soluble antibodies, consistent with prior literature [51]. Notably, inclusion of the costimulatory antibodies αCD2, αCD5, and αCD28 produced the highest levels of activation due to signaling of known co-stimulatory receptors [20,27,33]. While LNPs that delivered to Jurkat cells effectively also induced high activation, activation was not directly correlated with mRNA translation. Still, this supports the importance of T cell activation with increased levels of transfection.

We also observed elevated binding and internalization of the tLNPs; notably, binding, internalization, and T cell activation were both increased in the tLNP-treated groups compared to controls (SI Appendix, Fig. S13). This observation is consistent with prior studies showing that engagement of the T cell receptor (TCR), together with co-stimulatory signals, can promote endocytic activity and downstream T cell activation [52-54]. Based on these results, the top-performing formulations were advanced for further evaluation in a human PBMC model.

3.3. Bispecific LNPs enhance transfection of T cells in a human mixed blood cell population through the specific engagement of antibody domains with their respective receptors on T cells

We next evaluated the capacity of bispecific LNPs to specifically transfect T cells in a human mixed blood cell population. LNPs modified with αCD3/αTCR, αCD3/αCD2, αCD3/αCD4, αCD3/αCD5, and αCD3/αCD28 were compared against unmodified LNPs and αCD3-LNPs. Each of these LNP groups were formulated encapsulating mRNA encoding mCherry. Peripheral blood mononuclear cells (PBMCs) were dosed with LNPs at 2000 ng mRNA/300,000 cells and incubated for 24 h. It was confirmed that at this dose LNPs were not toxic to the PBMCs (SI Appendix, Fig. S14). Using flow cytometry, we evaluated mCherry expression in T cells (CD3⁺, CD4⁺ helper T cells, and CD8⁺ cytotoxic T cells), CD19⁺ B cells, and CD14⁺ monocytes (SI Appendix, Fig. S15).

When examining CD4⁺ helper T cells, we observed that bispecific LNPs modified with αCD3/αCD4 achieved the highest transfection efficiency (~8%) compared to nanoparticles lacking the αCD4 targeting ligand (Fig. 3A), indicating the ability of αCD3/αCD4-LNPs to preferentially target the CD4⁺ T cell subpopulation. In contrast, CD8⁺ T cells showed no significant differences in transfection efficiency following incubation with the various LNP formulations and exhibited overall lower transfection levels than those observed in the CD4⁺ subset (Fig. 3B). For CD3⁺ T cells, treatment with the monospecific αCD3-LNPs led to transfection of approximately 2.9% of CD3⁺ cells, while the addition of bispecific LNPS containing αCD3 and another antibody (αTCR, αCD2, αCD4, αCD5, and αCD28) led to increased transfection in all groups except for the αCD3/αCD4-LNPs (Fig. 3C). These findings suggest that the enhanced transfection observed with bispecific tLNPs may be driven by increased targeting of CD4⁺ T cells through CD4-directed ligands, as well as engagement of costimulatory receptors such as CD2 and CD28. We note that an isotype-LNP control was not included in the PBMC targeting experiments. While such a control could provide an additional assessment of non-specific antibody-mediated nanoparticle interactions, the differential association observed between tLNPs and untargeted LNP formulations supports receptor-dependent targeting under the conditions tested. Future studies incorporating isotype-conjugated LNP controls and receptor-blocking approaches would further strengthen evaluation of targeting specificity in primary PBMCs.

Fig. 3.

Bispecific LNPs enhance transfection of primary T cells in a human mixed blood cell population in vitro. (Iso, isotype control; TCR, αCD3/αTCR; CD2, αCD3/αCD2; CD4, αCD3/αCD4). Flow cytometry reveals mCherry⁺ CD4⁺ (a), mCherry⁺ CD8a⁺, and (b) mCherry⁺ CD3⁺ (c) T cells as a percentage of CD4⁺, CD8⁺, and CD3⁺ T cells respectively 24 h after human PBMCs are treated with PBS or mCherry-mRNA containing unmodified LNP, or tLNPs at a dose of 2000 ng/300,000 cells. (d) CD3⁺ T cells as a percentage of total alive cells in human PBMCs 24 h after treatment. mCherry⁺ CD19⁺ cells (e) and mCherry⁺ CD14⁺ cells as a percentage of CD19⁺ (b cell marker) and CD14⁺ (macrophage cell marker) respectively 24 h after human PBMCs treatment. For (a-c,e) n = 4 biological replicates and for (d,f) n = 3 biological replicates, error bars = standard error of the mean. Statistical analysis included RM one-way ANOVA with multiple comparisons, *p < 0.05 as compared to αCD3-LNPs (a-c, e) and as compared to LNP (d, f).

We observed a decrease in CD3⁺ staining when T cells were incubated with tLNPs (Fig. 3D). In control groups treated with PBS or unmodified LNPs, approximately 65% of PBMCs were CD3⁺. In contrast, following treatment with modified LNPs, detectable CD3 decreased to approximately 3% of total cells. This substantial reduction in CD3 signal suggests specific antibody–receptor engagement between tLNPs and CD3 at the cell surface. The diminished CD3 staining may result from receptor internalization following ligand binding or from surface receptor occupancy by tLNPs, thereby limiting antibody accessibility during staining [52,54,55].

We next evaluated off target uptake by other cell types, which is a concern in the design of targeted delivery vehicles. Both CD19⁺ B cells and CD14⁺ monocytes were evaluated to analyze the potential for off target transfection of the tLNPs. All LNPs resulted in low mCherry expression in CD19⁺ cells, however, the highest mCherrry expression occurred when using αCD3/αCD5-LNPs with ~1.5% of B cells transfected (Fig. 3E). This may be attributed to the dim expression of the CD5 scavenger receptor that has been shown to exist on human B cells and is often used to identify B1 and B2 cell subsets [56]. The average monocyte transfection efficiency decreased substantially when using tLNPs compared to untargeted LNPs, dropping from approximately 57% to 20% across all antibody-conjugated LNP groups (Fig. 3F). Although tLNPs still exhibited moderate transfection efficiency in monocytes, around 20%, which remains relatively high compared to T cells, this reduction is a potential indicator of increased selectivity or an effect of the DNA-linker. Further studies would be necessary to establish this mechanism. Since monocytes and macrophages are major contributors to nonspecific nanoparticle clearance, high uptake by these cells is typically undesirable when the goal is to deliver mRNA to other target cell types, such as T cells. Our findings demonstrate that incorporating T cell-specific targeting ligands onto LNPs not only enhances transfection of the intended cell population but also minimizes nonspecific uptake by non-target immune cells, thereby improving the delivery specificity of mRNA-loaded nanoparticles.

3.4. Antibody conjugated LNPs facilitate organ specific transfection in the spleen

Finally, we evaluated our bispecific LNPs in vivo in a mouse model. For this investigation, we used tLNPS that resulted in the highest transfection of T cells in vitro: LNPs conjugated with αCD3/TCR, αCD3/αCD2, and αCD3/αCD4-LNPs, and compared them to control groups including PBS, unmodified LNPs, an isotype control (isoLNPs), and single antibody αCD3-LNPs. All LNPs contained a mRNA encoding Fluc and each of the LNP groups were administered to mice at a dose of 0.6 mg/kg of body weight. At time points of 0.5 h, 5 h, and 24 h images were taken using the IVIS to capture luminescent expression over time indicating functional delivery of mRNA and at 24 h organs were collected and imaged individually (Fig. 4 A and SI Appendix, Fig. S16).

Fig. 4.

tLNPs mediate spleen transfection and induce T cell response. (Iso, isotype control; TCR, αCD3/αTCR; CD2, αCD3/αCD2; CD4, αCD3/αCD4). (a) Schematic of workflow for studying biodistribution of tLNPs. (b) Representative IVIS images of organs harvested from mice 24 h after intravenous injection of PBS, Isotype, and αCD3/αTCR-LNPs containing Fluc mRNA at a dose of 0.6 mg/kg body weight. (c) Measurements of luminescence for the spleen and liver as regions of interest on the IVIS (n ≥ 3 biological replicates, error bars = standard error of the mean, outlier removed using ROUT Q = 1%) and ratio of the luminescence for the spleen and liver of each mouse. Statistical analysis included ordinary one-way ANOVA with Dunnett’s multiple comparisons test, *p < 0.05 as compared to αCD3-LNP. (d) Quantification of histological sections of the spleen harvested 24 h after intravenous injection stained with H&E, αCD4, and αCD8 (n = 3 ROI) Statistical analysis included ordinary one-way ANOVA with Dunnett’s multiple comparisons test, *p < 0.05 as compared to isotype-LNPs. (e) Spleen weight normalized to mouse body weight 24 h after treatment. Statistical analysis included ordinary one-way ANOVA with Dunnett’s multiple comparisons test *p < 0.05 and **p < 0.01 as compared to isotype-LNP.

We first evaluated the organ specific transfection of tLNPs, focusing on the liver and spleen, but also evaluated other organs (SI Appendix, Fig. S17 and Fig. S18). The images of luminescence from the spleen and liver revealed differences in the biodistribution of the tLNPs as a function of their surface targeting (Fig. 4B). After selecting a consistent region of interest and normalizing to the background, the flux was calculated revealing that αCD3/αTCR-LNPs had the highest transfection in the liver and in the spleen (Fig. 4C). tLNPs shifted transfection towards the spleen from the liver relative to untargeted LNPs. The highest spleen to liver luminescence ratio occurred upon administration of the αCD3-LNPs. This ratio declined for LNPs targeted with αCD3/TCR and αCD3/αCD2-LNP because although there was higher luminescence observed in the spleen, there was also much greater luminescence in the liver (Fig. 4C).

αCD3/αCD4-LNP had the lowest luminescence observed in the spleen of all tLNP groups possibly due to targeting of a subset of T cells rather than the entire T cell population. In contrast, untargeted LNPs predominantly transfected the liver, leading to the lowest spleen:liver luminescence ratio of all formulations (Fig. 4E). This aligns with previous work which has shown that Dlin-MC3-DMA LNP’s tend to accumulate in the liver [43,57,58].

We propose two potential explanations for the decline of spleen:liver ratio with bispecific LNPs. First, the bispecific LNPs may exhibit increased off-target interactions, leading to enhanced accumulation in both the spleen and liver, but disproportionately in the liver. This would suggest reduced targeting precision compared to the monospecific αCD3-LNP. Second, liver sinusoidal endothelial cells and Kupffer cells express Fc receptors (FcRs) that mediate clearance of IgG-containing complexes [59-61]. We propose that the inclusion of two distinct antibodies on bispecific LNPs introduces a broader diversity of Fc-associated epitopes compared to a single-antibody formulation. This increased epitope heterogeneity may enhance recognition by complement proteins and Fc receptor–expressing cells, thereby promoting opsonization and FcR-mediated uptake in the liver. Thus, while bispecific targeting may improve avidity towards the intended T cell populations, it may simultaneously increase Fc-dependent hepatic sequestration through enhanced complement-mediated opsonization and clearance by the reticuloendothelial system. This highlights the importance of considering Fc-mediated clearance in nanoparticle design and supports the potential use of Fc-free formats, such as Fab fragments or single-chain variable fragments, to reduce unintended liver accumulation and improve targeting specificity.

At 24 h the physical characteristics of the spleen were analyzed for spatial distribution of T cells and physical enlargement because the white pulp of the spleen is a major site for T cell activation and proliferation [62,63]. Histological sections of the spleen were stained using hematoxylin and eosin (H&E), and CD4 and CD8 markers (Fig. 4D and SI Appendix, Fig. S19). H&E revealed expanded of areas of white pulp for αCD3/αTCR-LNPs as compared to the control groups (Fig. 4F). The CD4⁺ and CD8⁺ cells appeared to remain concentrated in the periarterial lymphatic sheath for all groups however there was expansion in these populations in αCD3/αTCR-LNP and αCD3/αCD2-LNP groups in comparison to the isotype control groups indicating that these cells were encountering T cell specific antibodies resulting in increased T cell production within the spleen. T cell proliferation generally follows T cell activation indicating an immune response triggered by tLNPs [17]. Notably, the T cell proliferation observed in the spleen was not seen in vitro Jurkat experiments or in PBMC viability likely because these were absent of key proliferative cues present in vivo, including interactions with antigen-presenting cells and the local IL-2–rich microenvironment, both of which are critical for sustaining T cell expansion [64,65]. Proliferation was not seen in Iso-LNP treatment indicating that the proliferation is due to the T cell specific ligands and not due to the LNP, mRNA, DNA-tethers, or overall antibody structure. In addition to this increase in splenic T cells, the overall spleen mass was also found to increase with the addition of T cell specific antibodies as compared to the unmodified LNP group indicating enlargement attributed to αCD3 and bispecific surface modification (Fig. 4E). Overall, this shows that antibody surface modification can impact LNP tropism and splenic T cell activity, however, other factors such as LNP lipid composition should also be considered.

3.5. Select bispecific LNP formulations enhance T cell transfection and change cytokine expression in vivo

Next, we evaluated the cell-specific transfection of tLNPs in vivo. Here, we analyzed T cells (CD3⁺, CD4⁺, and CD8⁺) and myeloid cells (CD11b⁺) present in blood and spleen 6 h after intravenous injection (Fig. 5A). Although circulating LNPs are cleared within ~5 h (SI Appendix, Fig. S20), mCherry protein expression reflects successful mRNA delivery and translation, which is detectable several hours post-uptake. LNP formulations were assembled encapsulating mRNA encoding mCherry and the % of mCherry⁺ cells in the blood and spleen were analyzed using flow cytometry (SI Appendix, Fig. S21). In the blood, we observed all bispecific LNP formulations improved total T cell transfection (CD4⁺ and CD8⁺) relative to αCD3-LNP and all LNP controls. The αCD3/αTCR-LNP had the highest % of mCherry⁺ cells in circulating T cells with ~10% transfection efficiency relative to the monospecific αCD3-LNP (3% transfection) and LNP controls (< than 2% transfection, Fig. 5B). The improved transfection of T cells with bispecific LNPs may be due to the capacity of these LNPs to engage with the TCR complex at two different epitopes causing an increase in receptor mediated endocytosis [52-54]. In blood resident myeloid cells, we observed bispecific LNPs also resulted in increased transfection relative to the monospecific αCD3-LNP and control LNPs. We hypothesize this enhanced transfection of monocytes may be due to the presence of two distinct Fc regions on the bispecific LNP surface resulting in increased engagement with scavenger receptors on monocytes [66]. To mitigate this uptake, it is advisable to use antibody fragments (i.e. nanobodies or single-chain variable fragments) rather than full length IgGs in downstream tLNP formulations. In the spleen, DNA-modified LNPs led to higher transfection efficiencies than the unmodified LNP, suggesting that the modification chemistry may have some capacity to increase splenic T cell transfection (Fig. 5D). The isotype-LNP had the highest transfection, though not significantly higher than other groups. Higher transfection by the isotype-LNP could be attributed to differences in spleen resident T cells and the protein corona absorbed to the isotype-LNP surface although further studies would be necessary to determine the exact mechanism [62,67,68]. Transfection of spleen resident monocytes remained low for all groups (Fig. 5E). Overall, the enhanced transfection of circulating T cells with GFP at levels as high as 10% is similar to levels needed for in vivo gene delivery approaches like CAR T therapies and support the capacity of DNA-anchoring to assemble tLNPs with therapeutic potential [16].

Fig. 5.

Select LNP formulations enhance T cell transfection and decrease cytokine release. (Iso, isotype control; TCR, αCD3/αTCR; CD2, αCD3/αCD2; CD4, αCD3/αCD4). (a) Schematic of the workflow for in vivo cell transfection study. Flow cytometry data showing the percent of T cells and monocytes expressing mCherry in the blood (b-c) and spleen (d-e). Statistical analysis included ordinary one-way ANOVA multiple comparisons test, **p < 0.01 (n = 2-3 biological replicates, error bars = standard error of the mean, points were omitted due to insufficient acquisition for analysis). Concentration of IL-6 and TNF-α in the serum 6 h after injection was measured (e-f). Bilirubin levels 6 h after injection were also assessed (g). Statistical analysis included ordinary one-way ANOVA multiple comparisons test, *p < 0.05 as compared to LNP (n = 3 biological replicates, error bars = standard error of the mean).

In the blood and spleen, consistent with our previous in vitro observations in PBMCs, we observed a decrease in CD3⁺ T cell across all tLNP-treated groups (SI Appendix, Fig. S22). This response has been seen across many CD3-targeted nanoparticle studies and is reported to be transient though the exact mechanism has not been thoroughly explored [16,17,69]. The αCD3-tLNPs and αCD3/αTCR-tLNPs resulted in the greatest decrease in detectable of CD3⁺ cells in the blood while αCD3/αCD2-tLNPs and αCD3/αCD4-tLNPs had greater CD3⁺ cell counts.

To further characterize the immunological impact of these bispecific constructs, we evaluated systemic cytokine responses. Serum was collected to assess IL-6 and TNF-α levels, as elevations in these cytokines are commonly associated with adverse inflammatory events such as cytokine release syndrome (CRS, Fig. 5E-F) [70,71]. After 6 h, control groups (PBS, LNP, and isotype) showed minimal deviation from baseline cytokine levels, indicating that neither the LNP platform nor the DNA-tethering system alone elicited a significant inflammatory response. In contrast, αCD3 and αCD3/αCD4-tLNPs induced the highest levels of IL-6 and TNF-α, suggesting a potential for low-grade CRS in these groups [72]. Notably, cytokine production associated with αCD3/αCD2-tLNPs was comparatively lower, yet still increased mRNA delivery in the blood as compared to αCD3-tLNPs. This highlights that strategic pairing of αCD3 with alternative co-targets can modulate downstream inflammatory signaling.

Given the well-established preferential accumulation of LNPs in hepatocytes, we next assessed total bilirubin as a marker of hepatobiliary function. Consistent with prior reports of DLin-MC3–based LNP systems, we detected minimal hepatotoxicity in most groups [57]. Although an increase in bilirubin was detected in the naked LNP group compared to PBS controls, incorporation of the DNA-lipid and DNA-conjugated antibody (isotype-LNP) did not further elevate bilirubin relative to LNP alone. Interestingly, the αCD3/αTCR-tLNP group exhibited an increase in bilirubin compared to the LNP group, which may correspond to the enhanced hepatic transfection observed in Fig. 4D and warrants further mechanistic investigation. Bilirubin levels in all groups remained below 2.0 mg/dL.

Overall, these data suggest that the method of DNA conjugation does not induce significant systemic inflammatory activation or liver toxicity under the conditions tested. Our results demonstrate that bispecific antibody selection not only influences T cell dynamics but also provides a tunable mechanism to balance therapeutic engagement with mitigation of toxicity, underscoring an opportunity to rationally design safer targeted drug delivery vehicles.

4. Conclusion

Using DNA-tethered conjugation of full-length commercial antibodies to Dlin-MC3-DMA LNPs, we demonstrate that tLNPs increase the efficiency of in vivo T cell transfection through specific antibody–receptor interactions. Specifically, we found that bispecific LNPs increase T cell transfection over single antibody formulations in both ex vivo studies with PBMCs and in vivo in the blood, showing that cotargeting of receptors can be used to improve cell-selective mRNA delivery.

The DNA-tethering approach offers several advantages. The DNA is site-specifically conjugated to the antibody Fc region, ensuring correct antibody orientation on the nanoparticle surface. This strategy allows rapid testing with commercially available antibodies, avoiding the time-intensive protein engineering required by methods such as ASSET. Moreover, the DNA scaffold offers the potential for precise control over antibody stoichiometry and spacing between the ligand and particle by varying DNA sequence or length, features that can significantly influence delivery efficiency, although these aspects are not explored in the present study [68,73]. By combining specificity, modularity, and tunability, oligonucleotide-mediated assembly overcomes many limitations of traditional chemical conjugation methods and provides a versatile platform for designing tLNPs.

This work highlights the potential of targeting the TCR complex in addition to costimulatory molecules as a strategic approach to improving the efficiency of LNP mediated T cell transfection. Bispecific LNPs, which included TCR or CD3 targeting antibodies, were more potent transfection agents than LNPs which did not target the TCR. The bispecific αCD3/αTCR-tLNP achieved the highest transfection rates in circulating T cells (~10%) (Fig. 5A) and highest splenic transfection (Fig. 4C). This suggests that targeting two epitopes on the same receptor complex creates a synergistic effect, which can also influence where nanoparticles distribute in the body. This approach may come with drawbacks, however, as indicated by the increased CD3⁺ depletion observed when T cells are engaged only through the TCR complex and without co-stimulation. In such cases, T cells with low CD3⁺ expression can become functionally unresponsive although some reports have also shown the ability of T cells to recover CD3⁺ expression after complete clearance of αCD3 from systemic circulation [17,74]. Although outside the scope of the current project, further work should be done to understand how CD3 availability following tLNP treatment may affect subsequent T cell functional activation to offer a more effective strategy for in vivo gene delivery.

In our study, bispecific tLNPs, that included co-stimulatory molecules such as αCD3/αCD2-LNPs and αCD3/αCD4-LNPs, exhibited increased transfection in the blood compared to αCD3-LNPs. They also increased transfection in a specific T cell subset, in this case CD4⁺ helper T cells. In previous work, bispecific tLNPs designed to simultaneously engage CD3 and a co-stimulatory receptor—examples include CD3/CD28 [24] and CD3/CD7 [25] exhibited enhanced activity relative to single-targeted counterparts, a trend mirrored in our results. The choice of costimulatory molecule appears to influence both cell-specific mRNA delivery, cell activation capacity, and cytokine release.

Transfection of circulating T cells is important for in vivo cellular engineering and immune modulation. Here, we acheived transfection of ~10% of circulating T cells, within rangeneeded for a potential therapeutic effect, though further optimization of LNP composition and antibody format remains possible. Significantly, bispecific LNPs offer the opportunity to tune cytokine responses potentially mitigating adverse health effects like CRS which are associated with CAR T therapies. More importantly, the DNA-tethering approach streamlines conjugation without the limitations of traditional chemical methods, enabling rapid, high-throughput in vivo screening of new bispecific or multivalent formulations. We anticipate the ease of assembling multivalent tLNPs and nanoparticles with this approach will allow for rapid selection of nanoparticle targeting ligands that are disease and target cell specific. Furthermore, the modularity of this approach allows for exchange of targeting molecules, control of the stoichiometry of multiple antibodies, and precise adjustment of tether length. This plateform for design of tLNPs supports the exploration of diverse receptor combinations to optimize specificity, efficacy, and biodistribution for future antibody–LNP therapeutics.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.jconrel.2026.114885.

Data availability

Data will be made available on request.