Bipolar CD4-targeted dual-DARPin-55/57 lipid nanoparticle enables efficient CRISPR/Cas-mediated HIV-1 DNA excision and reactivation blockade in latent CD4 T cell lines

TL;DR

A new targeted delivery system uses lipid nanoparticles to efficiently remove HIV-1 DNA from infected cells and prevent its reactivation.

Contribution

A dual-DARPin-LNP platform enables precise CRISPR/Cas9 delivery to CD4 T cells for HIV-1 proviral excision and reactivation blockade.

Findings

Dual-DARPin-LNP co-delivered spCas9 mRNA and sgRNAs to excise HIV-1 proviral DNA in latent cell lines.

Provirial excision blocked HIV-1 reactivation after stimulation with latency-reversing agents.

LNPs remained stable for over a year at 4 °C, preserving mRNA integrity.

Abstract

The persistence of HIV-1 latent reservoirs remains the principal barrier to a cure, as viral rebound occurs upon interruption of antiretroviral therapy. CRISPR/Cas genome editing offers a promising strategy to excise proviruses from host genome; however, the absence of a targeted and clinically viable delivery platform has hindered its translational application. Here, we report a chemistry-driven, CD4-targeted lipid nanoparticle (LNP) delivery platform employing a unique bipolar conjugation strategy to decorate dual CD4-targeted Designed Ankyrin Repeat Proteins (DARPins-55 and -57) on LNP (dual-DARPin-LNP). The N- and C-terminally modified DARPin-55/57 was thiolated stepwise, then bipolar maleimide-thiol coupling conjugated the thiolates to the maleimide-functionalized LNP surface. This coupling strategy ensured DARPin proper orientation on the LNP surface for efficient uptake by resting CD4 T cells. This dual-DARPin-LNP system was engineered for selective and efficient co-delivery of spCas9-GFP mRNA (Sp9m) and HIV-1-specific single-guide RNAs (sgRNAs) targeting LTR and Gag (LGsg) into HIV-1 latently infected CD4 T cells. In widely used HIV-1 latency models with defined proviral modifications (J-Lat 10.6 and 2D10 cell lines), dual-DARPin-LNP loaded with Sp9m/LGsg efficiently excised integrated HIV-1 proviral DNA, as confirmed by standard PCR genotyping, absolute digital PCR quantification, confocal microscopy, and flow cytometry. Importantly, proviral excision functionally blocked HIV-1 reactivation following stimulation with latency-reversing agents suberoylanilide hydroxamic acid (SAHA) and TNFα. Together, these findings establish a modular, non-viral, receptor-guided delivery platform for CD4 T cell targeting and provide proof-of-concept for precise HIV-1 DNA excision and reactivation blockade in established latency models. This new strategy represents a step toward next-generation curative interventions against persistent HIV-1 infection.

Image 1

•First dual-DARPin LNP with orientation-specific surface anchoring for CD4 targeting.•Selectively delivers cargo to resting CD4 T cells via bipolar DARPin conjugation.•Co-delivery of spCas9 mRNA and LTR1/GagD sgRNAs ensures efficient editing.•Cold-chain-stable LNPs preserve mRNA integrity for over one year at 4 °C.•Effective proviral HIV-1 excision prevents reactivation in latent reservoirs.

First dual-DARPin LNP with orientation-specific surface anchoring for CD4 targeting.

Selectively delivers cargo to resting CD4 T cells via bipolar DARPin conjugation.

Co-delivery of spCas9 mRNA and LTR1/GagD sgRNAs ensures efficient editing.

Cold-chain-stable LNPs preserve mRNA integrity for over one year at 4 °C.

Effective proviral HIV-1 excision prevents reactivation in latent reservoirs.

Figures

Table 2Fluorochrome-conjugated antibodies used for PBMC staining.Table 2Target antigenFluorochromeClone (Ab type)Host/isotypeVendorCatalog numberCD8 (T cell marker)PERPA-T8 (monoclonal)Mouse IgG1, κBioLegend301007CD3 (pan-T cell marker)Brilliant Violet 421 (BV421)UCHT1 (monoclonal)Mouse IgG1, κBioLegend300434CD4 (T helper cell marker)Alexa Fluor 700 (AF700)RPA-T4 (monoclonal)Mouse IgG1, κBioLegend300526CD16 (Monocyte/NK marker)Brilliant Violet 785 (BV785)3G8 (monoclonal)Mouse IgG1, κBioLegend302048CD19 (B cell marker)PE-Cy7HIB19 (monoclonal)Mouse IgG1, κBioLegend302216Viability (Live/Dead)Zomb

Full text

Introduction

1

HIV-1 infection remains incurable in the era of antiretroviral treatment (ART) due to viral latency. Latently infected CD4 T cells and macrophages persist across many tissues, such as secondary lymphoid tissues, the gastrointestinal and genital tract, bone marrow, and brain. Cessation of ART quickly leads to viral rebound even after decades of continued viral suppression. As of 2024, there are over 40.8 million people with HIV-1 (PWH) worldwide, representing a sustained global public-health burden. Moreover, PWH even under long-term ART have an increased risk of developing various comorbidities, such as precocious aging, neurocognitive disorders, and cardiovascular diseases [[1], [2], [3]].

Accordingly, approaches that permanently silence or eliminate HIV-1 provirus in latently infected cells are urgently needed to achieve an HIV-1 cure. Strategies under active investigation include latency reversal coupled with immune-mediated clearance [[4], [5], [6], [7], [8], [9], [10]], proviral eradication with genome editing [[11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32]], and long-term “deep” silencing [[33], [34], [35], [36]]. The CRISPR/Cas genome editing system has been demonstrated as a promising strategy for curing HIV-1 by excising proviral DNA from infected cell genome [[11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32]] with or without modifying CCR5 co-receptor to render them resist to HIV-1 new infection. Despite its potential, several challenges hinder its clinical application, the major obstacle being the difficulty in targeting and delivering CRISPR/Cas to HIV-1 latently infected cells, primarily resting CD4 T cells that harbor transcriptionally silent proviral DNA. The underlying cause of this challenge is due to the distribution of latently infected CD4 T cells across many distinct anatomical sites with different accessibilities and a low frequency of latently infected cells estimated at 1-1000 per million CD4 T cells [[37], [38], [39], [40]]. Effective delivery systems are thus essential to reach these low frequency reservoir cells efficiently. Both viral vectors like lentivirus (LV) and lentivirus-like particles and non-viral systems such as nanoparticles, especially lipid nanoparticles (LNP), are promising gene and drug delivery platforms, each with distinct mechanistic strengths and limitations for immunology and gene-editing research [41]. The LNP system offers a biocompatible delivery method with high cargo capacity and has been successfully used in RNA-based therapies and vaccines [42]. LNPs loaded with CRISPR/Cas gene editors offer high payload capacities, often surpassing viral vectors while enabling rapid, scalable synthetic production [43]. The major advantages of LNPs as non-viral cargo-delivery system include the lower immunogenicity, avoidance of viral genome integration risks, and enhanced biocompatibility, which is critical for in vivo gene editing, cancer immunotherapy, and vaccine delivery (e.g. COVID-19 mRNA vaccines) [[43], [44], [45], [46], [47]]. Nonetheless, achieving efficient and cell-specific delivery of LNPs, particularly in quiescent or resting CD4 T cells, remains a significant challenge, in part due to limited basal endocytosis (cell targeting) and inefficient cytosolic release in these cells [48,49].

Another critical consideration is the design of single-guide RNAs (sgRNAs) that can effectively bind conserved regions of the HIV-1 genome to overcome extensive heterogeneity of HIV-1 variants. The virus's integration into diverse genomic loci and its rapid evolution, particularly in the envelope gene, complicate targeted excision. However, conserved sequences within the long terminal repeats (LTRs) and structural genes such as Gag and Pol offer viable targets. Previous studies have demonstrated that CRISPR/Cas9 systems targeting multiple sites within the LTR can excise proviral DNA without harming host cells and can prevent reinfection [11,13,14,17,18,[25], [26], [27]]. Multiplexed sgRNA strategies targeting both LTRs and structural genes have shown enhanced efficacy in reducing viral replication and may offer a more robust approach [18,24,25,50].

To address these challenges, we developed a novel CD4-targeted LNP system for multiplex RNA cargo delivery, co-delivering spCas9-GFP mRNA (Sp9m) and a combination of LTR1-and GagD-targeting sgRNAs (LGsg) [11,18,[25], [26], [27]]. This system was designed with several strategic features: (1) delivery of spCas9 mRNA instead of protein to extend the editing window, reduce off-target effects, and facilitate GMP-compliant production [51]; (2) inclusion of sgRNAs targeting conserved regions in the LTR and Gag to minimize viral escape [18,24,25,50]; (3) use of LNPs for efficient and scalable delivery [[52], [53], [54]]; and (4) surface functionalization of LNPs with human CD4-specific Designed Ankyrin Repeat Proteins (DARPins) to ensure selective and efficient targeting of resting CD4 T cells [[55], [56], [57], [58], [59]].

DARPins, as targeting molecules, offer high specificity and affinity for the CD4 receptor while remaining immunologically inert and highly stable in preclinical models [57,[60], [61], [62], [63]]. Notably, CD4-binding DARPins can compete with HIV-1 gp120, thereby blocking infection by both R5 and X4 viral strains [60,64]. DARPin-decorated LVs have achieved >95% receptor-specific transduction of defined T-cell subsets in preclinical studies [65,66]. Further, DARPin-LV-like particles deliver mRNA with high selectivity and transgene expression in human CD8 T cells with markedly reduced off-target uptake [[65], [66], [67], [68], [69], [70], [71]]. However, effective targeting specificity and affinity of DARPin-conjugated LNPs can only be achieved when the targeting molecules are properly displayed on the LNP surface with their ligand-binding domains oriented outward to enable interaction with the CD4 receptor. In this study, we developed a bipolar, orientation-specific conjugation strategy that anchors dual DARPin-55/57 (DR) molecules via both N- and C-terminal linkages on the LNP surface. This modular approach can be readily adapted to display other targeting ligands through similar chemical modifications.

Our CD4-targeted DR-LNPs demonstrated selective uptake by resting CD4 T cells in peripheral blood mononuclear cells (PBMCs) from healthy donors, with minimal off-target accumulation. In latent HIV-1 model systems (J-Lat 10.6 and 2D10), these LNPs efficiently mediated proviral excision and reactivation blockade. Collectively, these findings establish CD4-targeted DR-LNPs carrying Sp9m and LGsg as a promising, cell-selective platform for excising HIV-1 proviral DNA. This study is intended as a delivery-platform proof-of-concept in established latency cell line models and is not designed as a comprehensive evaluation in primary patient-derived reservoirs or in vivo systems.

Methods

2

The lipids

2.1

1,2-Distearoyl-sn-glycero-3-phosphocholine (DSPC; 99% purity) was purchased from ABP Biosciences (Beltsville, MD, USA). SM-102 (99% purity) was obtained from DC Chemicals (Shanghai, China). 1,2-Dimyristoyl-rac-glycero-3-methoxypolyethylene glycol-2000 (DMG-PEG2000; 99% purity) was purchased from BroadPharma (San Diego, CA, USA). DSPE-PEG(2000)-maleimide and cholesterol (>99% purity) were obtained from Sigma-Aldrich (St. Louis, MO, USA).

DR expression and purification

2.2

Codon-optimized cDNAs encoding DARPin-55, DARPin-57, or the DR fusion protein (connected by a GS linker) were synthesized by BioMARTIK (Ontario, Canada). Inserts were cloned into pET30a vector using the NEB HiFi DNA Assembly Kit and transformed into E. coli DH5α. Positive clones were confirmed by diagnostic restriction digestion and Sanger sequencing. Verified plasmids were transformed into One Shot BL21(DE3)pLysS chemically competent cells (Thermo Fisher Scientific), and single colonies were expanded to generate glycerol stocks stored at −80 °C. For protein expression, single colonies were inoculated into 5 mL 2 × YT medium containing kanamycin (50 μg/mL) and grown overnight at 37 °C. Overnight cultures were used to inoculate 400 mL fresh 2 × YT medium containing kanamycin (50 μg/mL) and grown at 37 °C with shaking (250 rpm) to an OD of ∼0.6. Expression was induced with 1 mM IPTG, and cultures were incubated at 28 °C for 18 h. Cells were harvested by centrifugation (8000×g, 15 min; Eppendorf 5910R), resuspended in 40 mL binding buffer (50 mM Tris–HCl, pH 7.5; 500 mM NaCl; 20 mM imidazole), and lysed by sonication (Qsonica 500; two cycles of 2 min, 5 s on/5 s off, 50% amplitude). Lysates were clarified by centrifugation (12,000×g, 30 min, 4 °C). The supernatant was applied by gravity flow to a Ni-Sepharose 6FF column (GE Healthcare; Cat. #17-5318-01). After collection of the flow-through, the column was washed with five bed volumes of binding buffer. Proteins were eluted using a stepwise imidazole gradient (50–500 mM) in binding buffer, and 1 mL fractions were collected. Protein concentrations were measured using the Bradford assay (Bio-Rad). For SDS–PAGE, 20 μL samples from lysate, flow-through, wash, and elution fractions were mixed with 4 × Laemmli buffer, heated to 98 °C for 5 min, resolved by SDS–PAGE, and stained with Coomassie Brilliant Blue (Fig. 1).Fig. 1Expression, purification, and functional validation of dual DARPin-55/57 (DR) targeting human CD4 T cells.(A) Amino-acid sequence and domain schematic of DR, including the N-terminal 10 × His tag and C-terminal cysteine motif used for thiol-based conjugation to LNPs. A representative structural model depicts DR (green) displayed on the LNP surface via bipolar attachment, with CD4 (brown ribbon) bound to the DARPin. (B) Workflow for DR bacterial expression and affinity purification, with purity assessment by SDS–PAGE. (C) SDS–PAGE of purified DARPin-55 and DR proteins. (D) Immunocytochemical verification of DR binding to human resting CD4 T cells using anti-His detection. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)Fig. 1

In vitro transcription of spCas9-GFP mRNA (Sp9m) and HIV-1-targeted sgRNAs

2.3

The plasmid TP581 (pMJ920), containing T7 and SP6 promoters, was obtained from Addgene (#42234; gift from Jennifer Doudna). Plasmid DNA was linearized with XbaI and purified by agarose gel extraction; the linearized plasmid was used as template for in vitro transcription of Sp9m. The sgRNA transcription templates were generated by PCR using plasmids TP864 (LTR1) and TP902 (GagD) as templates [11,18]. The forward primer T403 (T7-U6/F; 5′-TTAATACGACTCACTATAGGGAGATTATATATCTTGTGGAAAGG ACG-3′) introduced a T7 promoter upstream of the sgRNA sequence, and the reverse primer T404 (gRNA/R; 5′-AAAAGCACCGACTCGGTGCCAC-3′) annealed downstream of the sgRNA scaffold. PCR was performed using Phusion High-Fidelity PCR Master Mix (Thermo Fisher Scientific; F531), and amplicons were purified using the QIAquick Gel Extraction Kit (Qiagen). Purified PCR products were used as templates for in vitro transcription of LGsg. Sp9m and LGsg were synthesized using the HiScribe T7 ARCA mRNA Kit (NEB; E2060) with N1-methyl-pseudouridine (TriLink BioTechnologies; N-1081). For mRNA, co-transcriptional capping with m7G anti-reverse cap analog (ARCA; Cat. #1411) and poly(A) tailing were performed according to the manufacturer's protocol. Template DNA was removed using Turbo DNase (Thermo Fisher Scientific), and RNA was purified using the Monarch RNA Cleanup Kit (NEB; E2040). RNA quantity and quality were assessed by denaturing Agarose gel and absorbance at 260 nm (with 260/280 ratio ≥1.8), respectively as we previously described [72].

To evaluate mRNA and sgRNA are efficiently loaded in the same LNP, we encapsulated Cy5-tagged mRNA (Cy5-Sp9m). The Cy5-Sp9m was synthesized by in vitro transcription as mentioned above. To fluorescently label the mRNA, Cy5-conjugated uridine triphosphate (Cy5-UTP; Jena Bioscience, Jena, Germany) was incorporated into the transcription reaction by partially substituting unlabeled UTP at an optimized molar ratio, as specified by manufacturer's protocol.

All steps were performed under RNase-free conditions using RNase-free reagents. Purified RNA was stored at −80 °C until use.

Generation and quality control of DARPin-LNP encapsulating Sp9m and sgRNAs

2.4

Formulation of cold-chain stable Sp9m LNPs with or without sgRNAs

2.4.1

Sp9m-loaded LNPs (Sp9m LNPs) and co-loaded Sp9m/LGsg LNPs were prepared using a modified rapid solvent exchange method previously established to enhance cold-chain stability [72]. Briefly, SM-102, cholesterol, DSPC, DMG-PEG2000, and DSPE-PEG(2000)-maleimide were dissolved in ethanol at a molar ratio of 50:38:10:1:1, respectively, with total lipid concentration maintained at 20 mM. The aqueous phase contained Sp9m with or without sgRNAs (LTR1 and GagD sgRNAs mixed 1:1, w/w) at a 1:1 (Sp9m:total sgRNA) weight ratio in 10 mM sodium citrate buffer (pH 4.0), prepared under RNase-free conditions. The ethanol phase was added dropwise to the aqueous phase at a 1:4 (v/v) ratio under constant agitation at 4 °C, with an N/P ratio of 8 to maximize encapsulation and minimize extracellular vesicle–mediated mRNA loss as previously reported [72]. The Cy5-tagged Sp9m/LGsg LNPs (Cy5-Sp9m/LGsg LNPs) were formulated alongside, to evaluate and ensure simultaneous loading of both mRNA and sgRNA within the LNPs. The resulting suspension was mixed for 12 h at 4 °C, followed by ethanol evaporation (“hardening”) under sterile conditions for ∼12 h at 4 °C. To enhance cold-chain stability, LNPs were dialyzed overnight at 4 °C against modified PBS (pH 7.4) supplemented with 50 μM N-tert-butylhydroxylamine (NtBHA) and 10 μM EDTA using 10 kDa MWCO dialysis cassettes with buffer exchange every 6 h. LNPs were stored at 4 °C; stability and retained activity were confirmed over long-term storage. Total encapsulated RNA was quantified using the RiboGreen assay as described below.

DARPin conjugation on Sp9m/LGsg LNPs

2.4.2

For CD4-targeted delivery, Sp9m/LGsg LNPs were surface-decorated with N- and C-terminal–modified DR using a three-step conjugation strategy (Fig. 2). First, DR proteins were thiolated at both termini via iminothiolation at pH 4.0 to introduce reactive sulfhydryls. Second, LNP pH was adjusted from 4.0 to 6.5 by dialysis against 1 × PBS to activate maleimide functionality on the LNP surface. Third, thiolated DR was added dropwise under gentle mixing to allow covalent maleimide–thiol coupling, generating a bipolar, stable thioether linkage on the LNP surface (Fig. 2A–C). After conjugation, DR-Sp9m/LGsg LNPs were dialyzed overnight at 4 °C against modified PBS using 100 kDa MWCO dialysis cassettes with buffer exchange every 6 h and stored at 4 °C until use. The DARPin conjugation efficiency on the surface of LNPs was evaluated using a bicinchoninic acid (BCA) protein assay according to the manufacturer's instructions. Briefly, LNP samples before and after surface conjugation were incubated with BCA working reagent, and absorbance was measured spectrophotometrically at 562 nm. Protein concentrations were quantified using a known concentration of DARPin-based standard curve, and conjugation efficiency was calculated by comparing the amount of conjugated DARPin normalized against associated with LNPs relative to input DARPin or. unconjugated LNP controls.Fig. 2Design and physicochemical characterization of CD4-targeted CRISPR/Cas9-loaded LNPs.(A) Schematic of LNP formulation encapsulating spCas9 mRNA (Sp9m) and HIV-1-targeting LTR1 and GagD sgRNAs (LGsg), followed by surface functionalization with CD4-binding DR to generate DR-Sp9m/LGsg LNP. (B) DR thiolation using Traut's reagent (2-iminothiolane) to introduce reactive sulfhydryls for site-directed conjugation. (C) Maleimide–thiol coupling of thiolated DR to maleimide-PEG–modified LNPs, yielding bipolar attachment while preserving CD4-binding activity. A representative structural model depicts CD4 (brown ribbon) bound to DR (green), illustrating an outward-facing presentation of the CD4-binding interface on DR-decorated LNPs after conjugation. (D) Particle size, concentration, polydispersity index (PDI), and ζ-potential of uncoated Sp9m/LGsg LNPs and DARPin-coated DR-Sp9m/LGsg LNPs. Data are mean ± SEM from six independent Sp9m/LGsg batches and five DR-Sp9m/LGsg batches. Statistics: two-way ANOVA with Šídák multiple comparisons (∗p < 0.05, ∗∗∗p < 0.001). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)Fig. 2

Particle size, polydispersity, and surface charge analysis

2.4.3

Hydrodynamic diameter, polydispersity index (PDI), and particle concentration were measured using a NanoSight NS300 (Malvern Panalytical, UK). Samples were diluted 1:200 in PBS to achieve an optimal particle count. The zeta potential and PDI were measured by dynamic light scattering (DLS) using a Zetasizer Nano ZS (Malvern Panalytical, UK).

Quantification of mRNA and encapsulation efficiency (EE)

2.4.4

Encapsulated Sp9m or Sp9m/LGsg was quantified using the Quant-iT RiboGreen RNA Assay Kit (Invitrogen). A standard curve was generated using serial dilutions of purified RNA (0–4000 ng/mL). LNP samples were diluted 1:50 and 1:100 to fall within the linear range. Empty LNPs were measured in parallel as a negative control. To release RNA, 150 μL of sample or standard was treated with 140 μL of 2% Triton X-100 and 10 μL heparin and incubated at 70 °C for 15 min. After cooling, 100 μL was transferred to a black 96-well plate and mixed with 100 μL RiboGreen reagent (1:100 in TE buffer). Fluorescence was measured at 485 nm excitation and 535 nm emission. RNA concentration was calculated from the standard curve (R^2^ = 0.99). Encapsulation efficiency (EE) was calculated as:

$$[eqn]$$

To evaluate % Sp9m loaded in Sp9m/LGsg LNPs, the Cy5-Sp9m/LGsg LNPs were first quantified using the Quant-iT RiboGreen RNA Assay Kit (Invitrogen). The same Cy5-Sp9m/LGsg LNPs sample solutions were then subjected to Cy5 fluorescence to verify and quantify Cy5-Sp9m in Cy5-Sp9m/LGsg LNPs. Cy5 fluorescence was measured using a SpectraMax® microplate reader (Molecular Devices, San Jose, CA, USA) at excitation/emission wavelengths of ∼650/670 nm. The Cy5-mRNA standard curve was generated using serial dilutions of purified Cy5-mRNA (0–4000 ng/mL). The % Cy5-Sp9m co-loading in Cy5-Sp9m/LGsg LNPs was calculated as (Table 1):

$$[eqn]$$

Table 1. Physicochemical characterization of LNPs.Table 1. TypeDynamic light scattering (DLS)Nanoparticle tracking analysis (NTS)RiboGreen assayBCA assayFluorescence assayPDISurface potential (mV)Size (nm)Nanoparticle concentration (NPs, e^12^/mL)%EEDARPin concentration (μg/mL)%Sp9m(Cy5) in total RNA loadSp9m/LGsg LNPs0.149 ± 0.01822.34 ± 0.7544.44 ± 12.73190.8 ± 6.2384.69 ± 4.68NDNDDR-Sp9m/LGsg LNPs0.167 ± 0.0103.89 ± 1.9795.20 ± 19.073.37 ± 0.7341.73 ± 3.9152.75 ± 6.0621.06 ± 0.87Data presented as mean ± standard error of means (SEM) of six different batches of CRISPR/Cas9 expressing LTR + GagD sgRNA/spCas9 mRNA loaded LNPs (Sp9m/LGsg LNPs; n = 6) and five different batches of CD4 targeting DARPin coated Sp9m/LGsg LNPs (DR-Sp9m/LGsg LNPs; n = 5). ND=Not determined; Sp9m(Cy5) = Cy5-tagged spCas9-GFP mRNA.

Physicochemical stability at 4 °C

2.4.5

To assess storage stability, six independent Sp9m LNP batches were stored at 4 °C (2 mL per batch). On day 0, 7, 15, 30, and 60, 200 μL aliquots were withdrawn and diluted 1:100 in PBS. Particle size distribution and concentration were measured by NanoSight NS300, while PDI and zeta potential were measured using the Zetasizer Nano ZS. Encapsulation efficiency was assessed by the RiboGreen assay as described above.

mRNA-LNP targeting in human PBMCs

2.4.6

PBMCs from three healthy donors were seeded at 1x10^5^ cells/well in round-bottom 96-well plates without activation and treated with DR-Sp9m LNPs (targeted) or Sp9m LNPs (non-targeted). Two independent LNP batches were tested. Cells were collected at 1, 4, 8, 16, 24, and 72 h post-treatment (Fig. 3). At each time point, cells were washed three times in FACS buffer (PBS with 1% BSA and 2 mM EDTA) and stained for 1 h at 4 °C with a multicolor antibody panel: PE anti-human CD8, BV421 anti-human CD3, Alexa Fluor 700 anti-human CD4, BV785 anti-human CD16, and PE-Cy7 anti-human CD19 (Table 2). Cells were washed three times, fixed in 1% paraformaldehyde overnight at 4 °C, washed, and analyzed on a CytoFLEX LX cytometer (Beckman Coulter) using CytExpert software. Data were analyzed in FlowJo v10.8.1. Gating included exclusion of debris, doublets, and dead cells (Zombie Red), followed by identification of lymphocyte/monocyte populations and marker-based discrimination of CD3/CD4/CD8/CD16/CD19 subsets (Fig. 3C). Cell viability was monitored by trypan blue exclusion, microscopy, and flow cytometry; no toxicity was observed at the effective dose.Fig. 3Formulation and cell-type targeting of DARPin-spCas9 mRNA LNPs in human PBMCs.(A, B) Physicochemical characterization of untargeted Sp9m LNPs and DARPin-decorated DR-Sp9m LNPs (size, PDI, ζ-potential, and particle concentration). Data are mean ± SEM from five independent Sp9m batches (n = 5). (C) Representative flow-cytometry gating strategy to quantify spCas9-GFP expression across PBMC subsets. (D) Time-course of spCas9-GFP expression in CD4 T cells, monocytes, B cells, and CD8 T cells after exposure to untargeted or DARPin-targeted LNPs (two independent LNP batches). (E) Uptake/expression kinetics in CD4 T cells and monocytes. Data are mean ± SE. Statistics: two-way ANOVA with Šídák multiple comparisons (∗p < 0.05, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001; ns, not significant).Fig. 3. Table 2Fluorochrome-conjugated antibodies used for PBMC staining.Table 2. Target antigenFluorochromeClone (Ab type)Host/isotypeVendorCatalog numberCD8 (T cell marker)PERPA-T8 (monoclonal)Mouse IgG1, κBioLegend301007CD3 (pan-T cell marker)Brilliant Violet 421 (BV421)UCHT1 (monoclonal)Mouse IgG1, κBioLegend300434CD4 (T helper cell marker)Alexa Fluor 700 (AF700)RPA-T4 (monoclonal)Mouse IgG1, κBioLegend300526CD16 (Monocyte/NK marker)Brilliant Violet 785 (BV785)3G8 (monoclonal)Mouse IgG1, κBioLegend302048CD19 (B cell marker)PE-Cy7HIB19 (monoclonal)Mouse IgG1, κBioLegend302216Viability (Live/Dead)Zombie Red™ (Fixable viability dye)BioLegend423109

HIV-1 proviral excision by DR-Sp9m/LGsg LNP in latent cell line models

2.5

Culture of HIV-1 latent cell lines and LNP treatment

2.5.1

The J-Lat clone (10.6), originally developed by Dr. Eric Verdin [73], was obtained from the NIH AIDS Reagent Program, Division of AIDS, NIAID, NIH. The 2D10 clone was donated by Dr. Jonathan Karn [74]. Both the J-Lat 10.6 and 2D10 cells were cultured in RPMI1640 containing 2.0 mM L-glutamine, 10% FBS and 1% penicillin/streptomycin. Both cell lines harbor HIV-1 provirus expressing GFP reporter that allows direct measurement and manipulation of HIV-1 latency and reactivation [75,76]. J-Lat 10.6 and 2D10 cells (1x10^5^ cells/well) were plated in flat-bottom 96-well plates and assigned to (i) untreated control (vehicle) or (ii) DR-Sp9m/LGsg LNP treatment. LNP treatment was performed using three independent LNP batches (n = 3 per batch). Approximately 3–4 h after plating (to allow equilibration), LNP-treated wells received DR-Sp9m/LGsg LNPs and control wells received an equal volume of 1 × PBS. Non-targeted LNPs were not used as excision controls because nonspecific uptake would still deliver Sp9m/LGsg, confounding interpretation of CD4-targeted delivery. Future mechanistic and safety studies in primary CD4 T cells and humanized mouse models will incorporate scrambled sgRNA controls to further assess specificity and off-target activity. After 72 h of treatment, both control and treated cells were stimulated with suberoylanilide hydroxamic acid (SAHA, 1 μM; Sigma-Aldrich) and recombinant human TNFα (10 ng/mL; PeproTech) for 24 h to induce proviral reactivation. Cells were then harvested and split: one portion for nucleic acid isolation (DNA/RNA) and the other for flow cytometry to assess GFP reactivation and cellular responses.

PCR genotyping and digital PCR analysis

2.5.2

Genomic DNA was extracted using the Quick-DNA Miniprep Plus Kit (Zymo Research; D7003) according to the manufacturer's protocol. DNA concentration and purity were measured using a Cytation 5 reader (Agilent), with absorbance at 260 nm and a A260/A280 ratio for purity. For PCR, 20 ng genomic DNA was amplified in a 12.5 μL reaction containing 2 × Phanta Max Master Mix (Vazyme), primers, and nuclease-free water. Cycling conditions were: 98 °C for 1 min; 35 cycles of 98 °C for 10 s, 60 °C for 8 s, and 72 °C for 8 s (primers T361/T458) or 72 °C for 1.5 min (primers V250/V251); final extension at 72 °C for 10 min. Products were resolved on 1.2% agarose gels and imaged on an iBright 1500 system (Thermo Fisher Scientific). Custom dPCR probes targeting the HIV-1 LTR, Ψ packaging signal, and RRE regions were designed (Table 3). Digital PCR was performed on the QuantStudio Absolute Q Digital PCR System (Thermo Fisher Scientific) using the program: 96 °C for 10 min; 40 cycles of 96 °C for 5 s and 60 °C for 15 s. Primer/probe sequences are listed in Table 3 and were validated in our prior studies [11,18,[25], [26], [27]]. Negative and positive controls were included in each run.Table 3. Oligonucleotides for sgRNAs targeting HIV-1-1 LTR, Gag and Pol, and PCR primers and probes.Table 3. Target NamePrimer name and DirectionSequenceGagT458: antisenseCCCACTGTGTTTAGCATGGTATTLTR-ET361: senseCACCGATCTGTGGATCTACCACACACAβ-actinV155: senseATCCTGCGTCTGGACCTGGCTGGCV156: antisensetACTCCTGCTTGCTGATCCACATCTGCJ-Lat-IT-FV250: senseCGTACTGGCTGGAGTAATAGCTTGCJ-Lat-IT-RV251: antisenseTTCCACAGTCTCATCCTTGCTACTGIPDA-Phi-FV10: senseCAGGACTCGGCTTGCTGAAGIPDA-Phi-ProbeV11: antisenseACTGGTGAGTACGCCIPDA-Phi-PCR-RT1845: antisenseGCACCCATCTCTCTCCTTCTAGCdPCR-HIV-1-LV-LTR-FV12: senseGCCTCAATAAAGCTTGCCdPCR-HIV-1-LV-LTR-RV13: antisenseGGCGCCACTGCTAGAGATTTTdPCR-HIV-1-LV-probeV14: ProbeAAGTAGTGTGTGCCCGTCTGTRRE-HIV-1-LV-FV15: senseAAACTCATTTGCACCACTGCRRE-HIV-1-LV-RV16: antisenseAATTTCTCTGTCCCACTCCATCRRE-HIV-1-LV-ProbeV17: ProbeTAGTTGGAGTAATAAATCTCTG

Confocal imaging of GFP dynamics

2.5.3

J-Lat 10.6 or 2D10 cells were treated with different concentration of DR-Sp9m/LGsg LNPs, while controls received an equal volume of 1 × PBS. After 72 h, cells were stimulated with SAHA (1 μM) and TNFα (10 ng/mL) for 24 h. GFP dynamics before and after stimulation were captured using a Stellaris 5 confocal microscope (Leica). Images were acquired using identical exposure and gain settings across conditions. Bright-field imaging was used to assess gross cell morphology/viability.

Quantification of GFP positivity by flow cytometry

2.5.4

Twenty-four hours after latency reactivation (SAHA/TNFα), cells from DR-Sp9m/LGsg LNP–treated and control groups were harvested and washed by centrifugation (300×g, 10 min). Supernatants were discarded and cell pellets were resuspended in 300 μL FACS buffer. Flow cytometry was performed on a BD LSRFortessa X-20 using BD FACSDiva v8.0. Data were analyzed in FlowJo v10.8.1. To quantify HIV-1 reactivation, GFP signal was measured in the FITC channel. A standardized gating workflow was applied to all samples: events were first gated on forward scatter area (FSC-A) versus side scatter area (SSC-A) to define the main cell population (“scatter” gate), followed by quantification of GFP-positive events within this parent population. Mean fluorescence intensity (MFI) of GFP was additionally calculated within the same gated population. To address potential concerns that LNP exposure could alter apparent cell size and bias scatter-based gating, FSC-A/SSC-A distributions were compared across treatment and stimulation conditions using singlet selection and a broadened cell gate; no FSC-A increase was observed following LNP treatment, and GFP + quantification was consistent across alternative gating approaches. Results are reported as % GFP+ (within the scatter-gated parent population) and GFP MFI.

Statistical analysis

2.6

Statistical analyses were conducted using GraphPad Prism version 10.4.2. The batch-to-batch formulation variability was analyzed by two-way ANOVA with Šídák's multiple comparison test. Comparisons between two independent groups used unpaired two-tailed t-tests. For ≥3 groups, one-way ANOVA with Tukey's multiple comparisons test was used where appropriate. The p-value below 0.05 was considered statistically significant, with significance levels indicated as follows: ‘∗’, ‘∗∗, ‘∗∗∗’, ‘∗∗∗∗’ indicates p < 0.05, 0.01, 0.001 and 0.0001, respectively. Data are expressed as mean ± standard error of the mean (SEM).

Results

3

Strategic design and purification of dual DARPin-55/57 (DR) for orientation-specific LNP surface loading

3.1

DARPins are small, single domain antibody mimetic peptides (14-15 kDa) that can bind to any given target protein with high affinity and specificity [77]. DARPin library screening has identified several DARPins that bind specifically to CD4-expressing human cells, especially the resting CD4 T cells, e.g., DARPin-55.2 and 57.2 [60]. Although individual CD4-targeted DARPin has shown to sufficiently mediate viral [66,[77], [78], [79], [80], [81], [82]] and non-viral [48] gene delivery, we hypothesized that dual DARPins may enhance CD4 receptor binding and maximize gene delivery in a synergistic or additive manner. To test this, we selected DARPin-55 and -57 because both showed highly specific binding to human CD4 receptors [60]. We used two residues: Glycine for flexibility, and serine for solubility to link DARPin-55 and -57 (DR) for moderate flexibility although other linkers can be used for further optimization. We further engineered DR with terminal modifications by introducing a 10 × histidine tag at the N terminus and 3x cysteine residues at the C terminus [83]. Prior to LNP loading, the N-terminal histidines and C-terminal cysteines were sequentially derivatized using a novel chemical modification strategy to generate N- and C-terminal thiol-functionalized DARPin molecules (Fig. 1A). The bipolar-end thiolation enabled directional and stable covalent conjugation of DR onto the LNP surface via thioether (-C-S-C- bond), ensuring that the CD4-binding site remained exposed following surface decoration (Fig. 1A). The DR fusion protein exhibited robust expression and high yield determined by affinity purification, comparable to individual DARPin-55 or -57 (Fig. 1B and C). The CD4 binding activity was validated by fluorescent labeling of purified DR and anti-His antibody immunostaining (Fig. 1D).

Design, formulation, and characterization of a novel DR-Sp9m/LGsg LNP formulation

3.2

The novel cold-chain-friendly DR-Sp9m/LGsg LNP was developed following our previous protocol with some modifications [72]. For a detailed protocol, see the Method section. This engineered CD4-targeted triple variable mRNA/sgRNA co-delivery system (DR-Sp9m/LGsg LNP) targets CD4-expressing cells ensuring co-delivery of all RNA components in the same cell for HIV-1 genome excision using CRISPR/Cas9 directed by dual LTR1 and GagD sgRNAs. (Fig. 2A). The Sp9m payload was utilized because 1) mRNA-mediated gene editing apparently outperforms ribonucleoprotein (RNP) and cDNA delivery [[84], [85], [86]]; 2) spCas9 remains the best editor for high efficiency both in vitro and in vivo, supported by a huge number of publications; 3) mRNA is functional in resting T cells in contrast to promoter-driven cDNA; and 4) GFP helps track delivery efficiency. The CRISPR/Cas9 excision was directed by dual sgRNAs targeting the LTR and structural region of the HIV-1 genome due to several advantages previously reported by our group [11,18,[25], [26], [27]] and others [13,14,17,24,[87], [88], [89], [90], [91]].

For DR-Sp9m/LGsg LNP formulation, a novel and modified formulation was adapted from our published formulation methodology [72]. The LNP formulation was adapted with a modified molar ratio, i.e., SM-102:Cholesterol:DSPC:DMG-PEG(2K) at 50:38:10:1 M ratio, respectively, along with the addition of a fifth lipid component, i.e., DSPE-PEG (2K)-Maleimide at 1 M ratio (Fig. 2A). The fifth lipid component was added to get Maleimide functional anger to conjugate DARPin molecules on the surface of the LNP. The rationale behind using DSPE over other lipid backbones was that saturated lipids (single bond) have closer packaging arrangements compared to lipids with unsaturated backbone [92]. The 1:1 M ratio of DMG-PEG(2K) to DSPE-PEG(2K)-maleimide was used to maintain optimal maleimide density without overcrowding on the LNP surface. The PEG(2K) spacer with DSPE (two stearic acid with 18 carbon (C18) tails [93] extended the maleimide group above the DMG-PEG layer (two myristic acid with 14 carbon (C14) tails [94], thereby preserving its accessibility for DARPin conjugation with minimal steric hindrance from neighboring PEG(2K) chains (Fig. 2A). To maximize mRNA/sgRNA loading, the N/P of 8 was employed to ensure efficient encapsulation and to minimize extracellular vesicle-mediated mRNA loss during cellular uptake.

Next, we sought to develop an mRNA–LNP formulation method with improved cold-chain stability. Unlike conventional formulations that require storage at −80 °C, our LNP-encapsulated mRNA remains stable for over one year at 4 °C, making it cold-chain compatible. To achieve this, we formulated Sp9m/LGsg LNP using our previously established modified mRNA–LNP formulation method [72]. Standard microfluidic LNP synthesis does not include an organic-phase evaporation or “hardening” step, which can result in loosely packed lipid layers and increase the likelihood of aqueous ion percolation into the nanoparticle core where mRNA is encapsulated. Interaction of mRNA with these infiltrating ions can lead to the formation of adducts and chemical impurities, thereby accelerating mRNA degradation and compromising nanoparticle stability. We introduced two key modifications to address this mRNA instability issue. First, immediately after LNP formation, we removed the organic phase by slow evaporation (Fig. 2A). This step enhanced lipid packing (“hardening”) and improved lipid-to-RNA encapsulation, reducing the chances of exposure of mRNA/sgRNA to the aqueous environment. Furthermore, traditional microfluidic LNPs often retain free RNA and lipids that generate degradation-related impurities over time through hydrolysis, oxidation, or reactions involving ionizable lipids, PEG-lipids, and RNA adducts (e.g., aldehydes, peroxides). These impurities can diffuse into the LNP core, allowing buffer ions to contact the encapsulated RNA [95]. As a result, many reported mRNA–LNPs lack long-term stability at 4 °C or even at −80 °C despite use of cryoprotectants [96]. To overcome this, we added a second modification: a post-hardening dialysis step. Following organic phase removal, we eliminated free lipids and mRNA by dialyzing the LNPs overnight at 4 °C against modified PBS (pH 7.4) supplemented with 50 μM NtBHA and 10 μM EDTA, using 1000-fold excess volume. EDTA acted as a quencher, while NtBHA prevented impurity formation, specifically residual or time-dependent degradation products derived from ionizable lipids, PEG-lipids, and RNA adducts. The organic phase removal improved LNP maturation, compactness, and uniformity, as reflected by diameter and polydispersity index (PDI) measurements (Table 1, Fig. 2D). The subsequent dialysis step removed post-LNP synthesis impurities while the exchange buffer composition prevented overtime impurity generation. The hardening step prevented ion exchange with the encapsulated RNA, dialysis removed post-LNP formulation free molecules/impurities and the modified buffer removed impurities that accumulate over time. Together, these modifications enabled long-term stability of mRNA–LNPs at 4 °C over one year.

For targeting CD4-expressing cells, DARPin molecules were decorated on the surface of LNPs loaded with Sp9m/LGsg by adapting a novel methodology with three major steps. To achieve this, we used engineered DR with 10 × histidine tag at the N terminus and 3x cysteine residues at the C terminus. The first two steps involved a novel sequential thiolation strategy leading to stable N- and C-terminal thiol functionalization of DARPin molecules. First, DARPin activation step, i.e., iminothiolation of DARPin molecules to reduce the N-terminal His-tag and C-terminal cysteine to stable thiol (-SH) groups, at pH 6.5 (Fig. 2B), leading to the activated DARPin molecule with bipolar (N- and C-terminal) thiol groups. Second, the LNP buffer adjustment step, which involves buffer exchange by dialysis to get to LNPs at pH 6.5-7.5 necessary for functionalized DARPin molecule stability and maleimide-thiol reaction (Fig. 2A). Third, the maleimide-thiol reaction (Michael addition) step, the maleimide group on the LNP surface interacts with the N- and C-terminal thiol group of activated DARPin, conjugating DARPin orientation to keep the CD4 receptor binding site open, as desired (Fig. 2C). This bipolar thiolation of DARPin molecules enabled site-specific conjugation to maleimide-functionalized LNPs, ensuring stable thioether bond (-C-S-C-) formation between the N- and C-terminal DARPin molecules on LNP surface (very stable with one orientation conjugation) and the accessibility of CD4-binding domain for T cell targeting. The conjugation of DARPin molecules on LNP surface was confirmed by biophysical characterization and BCA protein assay. The increase in size and decrease in surface charge indicate physical binding of DARPin (Fig. 2D; Table 1). The BCA assay confirmed presence of DARPin on the Sp9m/LGsg LNP surface (Table 1). The non-significant change in PDI value indicates uniform DARPin molecule conjugation on the LNP surface without causing any aggregate formation (Fig. 2D). These findings, derived from six independent LNP batches (n = 6) and five DARPin-coated formulations, were statistically validated using two-way ANOVA with Šídák's post hoc test, confirming the reproducibility and stability of the targeted LNP system.

DARPin display on LNP improves targeted delivery to resting CD4 T cells

3.3

Precise, cell-specific delivery of CRISPR/Cas editors is essential to eliminate latent HIV-1 reservoirs that persist despite antiretroviral therapy. Successful excision of integrated HIV-1 provirus requires co-delivery of Cas9 and multiple sgRNAs targeting conserved regions of the viral genome in HIV-1-infected CD4 T cells. Targeting via the CD4 receptor increases both the precision and efficiency of proviral excision strategies [97,98]. To enable simultaneous delivery of both CRISPR components, we formulated a single DR-LNP co-loaded with Sp9m and HIV-targeting sgRNAs at defined ratio. To evaluate and ensure Sp9m and sgRNA co-loading in LNP, we co-loaded Cy5-labeled Sp9m and LGsg in LNP. Quantitative fluorescence and RNA loading analyses showed that Sp9m comprised ∼21% of the total nucleic acid payload, while sgRNAs accounted for ∼79% (Table 1). To evaluate the targeting efficacy of DARPin-decorated mRNA LNPs, we formulated LNPs loaded with mRNA expressing spCas9-GFP (DR-Sp9m LNPs). Five independent batches were produced and characterized biophysically. The DR-Sp9m LNPs demonstrated consistent particle size, polydispersity index (PDI), surface charge (ζ-potential), and particle concentration, confirming their stability and reproducibility (Fig. 3A and B). Targeting efficiency and cargo expression were evaluated using human PBMCs, which contain CD4-expressing immune subsets. Cells were treated with either DARPin-targeted Sp9m LNPs or untargeted Sp9m LNPs and analyzed over a period of 72 h. Flow cytometry revealed that DARPin-functionalized LNPs yielded a significantly higher proportion of GFP CD4 T cells compared to untargeted controls (Fig. 3C–E). Among all immune subsets evaluated, CD4 T cells and monocytes exhibited the highest early uptake (Fig. 3D), consistent with DARPin-binding to CD4 receptor. Although monocytes internalized more particles, GFP expression did not differ between targeted and untargeted groups at later time (Fig. 3E), likely due to nonspecific uptake via native phagocytosis rather than receptor-mediated delivery, which occurred earlier. These findings demonstrate that DARPin conjugation enhances selective uptake and functional mRNA expression in CD4 T cells, establishing a critical foundation for precise CRISPR-mediated gene editing in HIV-1-infected immune cells.

DARPin-LNP delivery of spCas9 mRNA and duplex LTR1/GagD sgRNA to HIV-1 latent J-Lat 10.6 cell line for effective deletion of HIV-1 proviral DNA

3.4

To evaluate the efficacy of DARPin-LNP-mediated delivery of CRISPR/Cas9 components in excising HIV-1 proviral DNA, we treated HIV-1 latent J-Lat 10.6 cell line with DR-Sp9m/LGsg LNPs. Regular PCR genotyping (Fig. 4A) revealed a marked reduction in HIV-1-specific amplicons in treated cells compared to untreated controls, indicating successful disruption of the proviral genome. This reduction was normalized against human β-actin level (Fig. 4A), confirming that the observed loss of HIV-1 signal was not due to cell loss. To further validate the excision of HIV-1 proviral genome, we designed primers targeting the host genome and flanking the full-length HIV/R7/E−/GFP genome (10,200 bp; heterozygous) integrated within intron 2 of the SEC16A gene on chromosome 9q34.3 [99], and performed PCR genotyping as we described previously [11,18,[25], [26], [27]]. Treatment with DR-Sp9m/LGsg-LNP completely eliminated the entire HIV-1-GFP genome in J-Lat 10.6 cell line (Fig. 4B). Surprisingly, we did not observe any excised fragmental bands that contain the expected 239-bp from the host genome and different viral sequences after LTR1/GagD editing as we reported previously [11,18,[25], [26], [27]]. This indicates that the “remaining” LTR after excision of entire HIV-1 genome may be cleaning out by the recombination of the host wild type allele genome.Fig. 4Elimination of HIV-1 Proviral DNA in J-Lat 10.6 Cell Line via DARPin-LNP Delivery of spCas9 mRNA (Sp9m) and LTR/GagD sgRNAs (LGsg).(A) Representative PCR gel images showing a significant reduction (80.6% ± 4.9% normalized to β-actin, n = 8) of the HIV-1 genomic region spanning from the 5′-LTR to Gag in J-Lat 10.6 cells treated with DR-Sp9m/LGsg LNPs. Data are from two independent experiments using two LNP batches (B5 and B7) at different dosages. (B) Complete excision of the full-length HIV-1-GFP genome integrated within intron 2 of the SEC16A gene. PCR was performed using primers (V250/V251) flanking the integration site at position 136,468,579 (GRCh38/hg38). (C–G) Digital PCR (dPCR) analysis quantifying HIV-1 proviral DNA, confirming a marked reduction in proviral copies following DR-Sp9m/LGsg LNP treatment. (C) Schematic of the HIV-1-GFP genome, dPCR probe locations, the excised region, and a representative 2D scatter plot distinguishing intact from excised genomes. (D, E) Representative 1D plots showing signal from individual probes and corresponding HIV-1 genome copy numbers. (F, G) Representative 2D plots from duplex probe assays, illustrating quantification of intact versus excised HIV-1 genomes.Fig. 4

To quantify the extent of proviral elimination, we performed digital PCR (dPCR) targeting conserved regions of the HIV-1 genome. As shown in Fig. 4C–G, cells treated with DR-Sp9m/LGsg LNP exhibited a dramatic and statistically significant decrease in HIV-1 DNA copy number detected by multiple probes targeting different regions of HIV-1 genome (Fig. 4C). This reduction was consistent across multiple replicates and independent experiments, demonstrating the robust and reproducible activity of the targeted LNP CRISPR system. These findings provide strong evidence that DARPin-LNP delivery enables efficient and specific genome editing in latently infected cells.

DARPin-LNP spCas9 treatment significantly blocked HIV-1 reactivation in J-Lat and 2D10 latency models

3.5

To evaluate whether CRISPR/Cas9–mediated disruption of latent HIV-1 proviral DNA prevents viral reactivation, J-Lat 10.6 cells were treated with DR-spCas9/LGsg LNPs, followed by stimulation with suberoylanilide hydroxamic acid (SAHA) and TNFα, two well-established latency-reversing agents [100]. As shown by confocal microscopy (Fig. 5A), LNP treatment resulted in a pronounced and dose-dependent reduction in GFP-positive cells compared with vehicle-treated controls, indicating effective suppression of HIV-1 reactivation. At the higher LNP dose, a modest reduction in cell density was observed, consistent with partial cytotoxicity. To quantify reactivation at the population level, we performed flow cytometry using a standardized gating strategy to identify GFP-expressing cells representing transcriptionally reactivated HIV-1. To address concerns that LNP binding might increase apparent cell size and thereby bias scatter-based gating, we compared FSC-A/SSC-A distributions across treatment groups. DARPin-LNP treatment did not increase FSC-A (no rightward shift), indicating no measurable increase in apparent cell size (Fig. 5B). Moreover, the GFP readout was consistent regardless of gating strategy: repeating the analysis with (i) singlet gating and (ii) broadened FSC/SSC inclusion gates yielded GFP-positive percentages that were essentially indistinguishable from the original gating analysis (Fig. 5C). Across six independent biological replicates (n = 6), LNP-treated cells exhibited a significant decrease in the percentage and mean fluorescence intensity of GFP-positive cells compared with control groups (Fig. 5D), confirming that spCas9-mediated proviral disruption robustly limits HIV-1 reactivation even under strong stimulatory conditions. To determine whether this inhibitory effect was conserved across latency models, we extended the analysis to 2D10 cell line. Consistent with the J-Lat 10.6 results, DR-Sp9m/LGsg LNP treatment markedly reduced GFP expression following reactivation stimuli (Fig. 5E), demonstrating effective blockade of HIV-1 reactivation in an independent latent HIV-1 system. Collectively, these data demonstrate that DARPin-LNP–mediated delivery of Sp9m/LGsg efficiently disrupts latent HIV-1 proviral DNA and functionally suppresses viral reactivation across multiple latency models. This approach significantly limits inducible HIV-1 transcription and supports the potential of targeted CRISPR/Cas9 LNP platforms as a promising strategy for reducing the inducible latent HIV-1 reservoir.Fig. 5DR-Sp9m/LGsg LNP suppresses latency reversal in J-Lat 10.6 and 2D10 models without FSC-based gating artifacts.(A) Representative confocal images of J-Lat 10.6 cells pretreated with vehicle or DR-Sp9m/LGsg LNPs (multiple doses), followed by latency reversal with SAHA (1 μM) + TNFα (10 ng/mL). Reactivation is reported by GFP expression, which is markedly reduced in LNP-treated cells. (B) Scatter/size-control analysis used to address potential LNP-associated shifts in apparent cell size: FSC-A distributions were compared across treatment/stimulation conditions after singlet selection and a basic cell gate (“Media FSC-A”), showing no FSC increase with LNP exposure (n = 3–6 samples/group). (C) Representative flow-cytometry gating strategy. Events were first gated on FSC-A versus SSC-A (“Scatter” parent gate), followed by GFP-A gating to quantify reactivated (GFP+) cells. Representative percentages are shown in the plots (e.g., Control + SAHA/TNFα: Scatter 18.4% with GFP+ 81.9% within the Scatter gate). (D) Quantification of latency reversal in J-Lat 10.6 cells, shown as (top) %GFP + cells and (bottom) GFP mean fluorescence intensity (MFI) across six independent biological replicates (n = 6), demonstrating strong SAHA/TNFα induction in controls and near-complete suppression after DR-Sp9m/LGsg LNP treatment. GFP + percentages are reported within the Scatter parent gate. Two-way ANOVA (treatment × stimulation) with Šídák multiple comparisons; ∗∗∗p < 0.001. (E) Representative confocal images of 2D10 cells ± DR-Sp9m/LGsg LNP pretreatment and ± SAHA/TNFα stimulation, showing concordant suppression of reactivation in an independent latency model.Fig. 5

Discussion

4

In this proof-of-concept study, we developed and validated a novel bipolar dual-DARPin-conjugated-LNP platform for the targeted delivery of CRISPR/Cas9 components to HIV-1 resting CD4 T cells, the primary reservoir of latent HIV-1 infection (Fig. 6). Our approach integrates several key innovations, including a modified cold-chain stable LNP formulation method, a creative bipolar conjugation strategy for stable and precise orientation of targeting molecules (DARPin) for enhanced delivery to resting CD4 cells, and the co-delivery of spCas9 mRNA with duplex sgRNAs targeting three conserved regions of the HIV-1 genome. Collectively, these advances support efficient and selective delivery in the models tested and provide a scalable framework for future evaluation in primary cells and in vivo HIV-1 reservoir models.Fig. 6**Model for dual DARPin-mediated LNP delivery of spCas9 mRNA and duplex HIV-1 sgRNAs to human CD4 T cells to enforce functional proviral excision.**A LNP incorporating spCas9 mRNA and duplex sgRNAs (targeting LTR1 and GagD) is functionalized with CD4-specific DARPin-55 and DARPin-57 molecules via bipolar thioether linkages. The dual-DARPin coating facilitates selective binding to CD4 T cells and receptor-mediated uptake. Following endosomal escape, Cas9/sgRNA complexes are assembled and execute targeted excision of integrated HIV-1 proviral DNA. Removal of the provirus abrogates HIV-1 reactivation upon latency-reversing stimulation, demonstrating a receptor-guided, non-viral strategy for HIV-1 proviral excision.Fig. 6

A major innovation of this work is the post-synthesis organic phase evaporation and purification process that follows LNP formation. Incorporating this step, together with dialysis to remove unencapsulated mRNA and free lipids, minimizes adduct and free-radical formation, thereby enhancing LNP maturation, compactness, and long-term stability. The use of a modified dialysis medium containing molecular quenchers to eliminate reactive ionic and lipid impurities further prolongs the stability of encapsulated mRNA. These optimizations yield highly uniform nanoparticles with consistent particle size, polydispersity index (PDI), and ζ-potential across multiple production batches. A second key innovation is the dual-site DARPin conjugation strategy, which ensures precise molecular orientation of the CD4-targeting DARPins on the LNP surface. By exploiting both N- and C-terminal thiol groups for site-specific conjugation, the DARPins are bipolarly anchored in a defined upright orientation, maximizing receptor accessibility while minimizing steric hindrance [101]. This bipolar configuration mimics the multivalent, geometrically ordered display of viral envelope proteins, resulting in enhanced binding avidity and selectivity toward CD4 T cells. This is consistent with previous report using bispecific CD4/CD32a DARPins decorating AAV2 [59]. Flow cytometry confirmed that DR-LNPs achieve substantially higher uptake in CD4 cells, with minimal off-target delivery to non-CD4-expressing cells, an essential property for reducing systemic exposure and mitigating CRISPR/Cas9 off-target activity. Previously reported single CD4-targeting DARPin delivery strategies using LNPs [48] or AAVs [66,[77], [78], [79], [80], [81], [82]] have demonstrated the feasibility of receptor-directed gene transfer. Our bipolar dual-anchor configuration and DR pairing introduces a distinct architectural strategy intended to enhance multivalent engagement and targeting precision. The present in vitro findings establish feasibility and provide a foundation for future quantitative benchmarking and in vivo validation. In addition to improved targeting, the resulting DARPin-anchored LNPs exhibit high molecular stability, reproducibility, and scalability, key prerequisites for clinical translation and cold-chain logistics. Unlike viral vectors that require complex and costly manufacturing pipelines, this modular LNP platform would allow rapid adaptation to diverse ligands (e.g. antibodies, nanobodies, peptides) and payloads as a conceptual design feature, with experimental demonstration reserved for future work, offering a streamlined, GMP-compliant, and clinically versatile system for therapeutic delivery. Future studies will determine whether this CD4-targeted bipolar DARPin architecture can also enhance AAV or virus-like particle–mediated gene delivery in human resting CD4 T cells.

Among the various strategies for Cas9 delivery of mRNA, cDNA, or RNP complexes, mRNA offers a compelling balance of efficiency, safety, and temporal control. Once in the cytoplasm, mRNA is rapidly translated and degraded, enabling a transient “hit-and-run” editing approach that minimizes off-target effects and avoids genomic integration risks associated with cDNA [[102], [103], [104]]. Moreover, mRNA allows for faster Cas9 expression and is amenable to chemical modifications (e.g., pseudouridine incorporation) to enhance stability and reduce immunogenicity. While RNPs offer immediate activity, they require complex protein purification and formulation. Importantly, mRNA is compatible with clinically validated delivery systems such as LNPs, which we employed in this study. Our data demonstrate that DARPin-mediated LNP targeting enables selective and functional transduction of CD4 T cells, but not monocytes, within heterogeneous PBMC populations. Although both CD4 T cells and monocytes exhibited early particle uptake, only CD4 T cells displayed a marked increase in GFP expression following treatment with DARPin-Sp9m LNPs. This finding indicates that DARPin conjugation promotes receptor-mediated delivery and productive cytosolic release of mRNA selectively in CD4 T cells, consistent with the DARPin's high-affinity binding to the CD4 receptor. In contrast, the lack of enhanced GFP expression in monocytes, despite their higher apparent particle uptake, suggests that internalization in these cells occurs primarily through nonspecific phagocytosis rather than receptor-dependent endocytosis. Such phagocytic uptake likely results in endosomal sequestration and degradation of LNPs, preventing efficient mRNA release and translation. These observations highlight an important mechanistic distinction between physical uptake and functional delivery. DARPin modification confers receptor-guided specificity that directs LNPs to CD4 T cells, facilitating efficient endosomal escape and cytoplasmic translation of mRNA cargo. This selective delivery minimizes off-target expression in non-CD4 populations and enhances the precision of CRISPR/Cas9-based genome editing in the intended target cells. Together, our findings underscore the value of DARPin-decorated LNPs as a rationally designed platform for cell-type–specific mRNA and gene-editing delivery, particularly for eliminating latent HIV-1 reservoirs residing within CD4 T cells. On the other hand, the higher nonspecific uptake of LNP-Cas9/sgRNA by myeloid cells may confer an additional benefit for HIV reservoir clearance, as monocytes also harbor proviral genomes [105,106]. We emphasize that cellular viability/toxicity readouts are not surrogates for genomic specificity; comprehensive off-target profiling will be undertaken separately using established sequencing-based assays.

Compared to AAV vectors, which are limited by packaging constraints, potential immunogenicity, and integration risks, DARPin-LNPs offer a safer and more controllable alternative. Notably, DR-LNPs significantly increase the proportion of CD4 T cells receiving both Cas9 and sgRNAs, ensuring co-delivery to the same HIV-1-infected cells, a critical requirement for effective proviral excision. The low immunogenicity of LNPs, combined with the engineered specificity of DARPins, supports the potential of this platform for repeated dosing and long-term therapeutic applications.

Targeting both the LTR and Gag regions of the HIV-1 genome enhances the likelihood of complete proviral excision, reducing the risk of viral escape and reactivation, as previously reported [11,18,24,27,[87], [88], [89], [90], [91]]. This multiplex targeting strategy is particularly important given the genetic diversity and integration patterns of HIV-1. The use of GFP-tagged spCas9 enabled real-time tracking of transgene expression and delivery efficiency. Digital PCR provided a highly sensitive and quantitative method for detecting residual HIV-1 DNA, confirming the efficacy of proviral excision, especially valuable for evaluating therapeutic outcomes in latent reservoirs with extremely low viral loads.

Efficient HIV-1 DNA excision in rare latently infected cells requires coordinated intracellular availability of both Cas9 and its sgRNAs. Our compositionally tuned co-loading strategy using Cy5-conjungated Sp9m addresses this requirement by enriching sgRNAs relative to mRNA, thereby ensuring immediate guide availability as Cas9 is translated and promoting rapid RNP assembly in situ. Notably, even though Sp9m represented only ∼21% of the nucleic acid payload, it was sufficient to drive robust spCas9 expression and downstream excision, while the excess sgRNA fraction likely increases the probability that newly translated Cas9 rapidly finds its guides. This co-delivery design may help maximize editing efficiency while minimizing total mRNA burden and supports the broader utility of single-particle mRNA/sgRNA co-formulation for CRISPR-based antiviral applications.

We demonstrated the efficacy of our DR-LNP platform in J-Lat cell line, achieving robust genome editing through the co-delivery of spCas9 mRNA and duplex sgRNAs targeting the HIV-1 LTR and Gag regions. This led to a significant reduction in proviral DNA, as confirmed by PCR genotyping and digital PCR quantification. The observed loss of HIV-1-specific amplicons, normalized to human β-actin, indicates that the reduction was due to specific cleavage and deletion of the proviral genome rather than cytotoxicity. Digital PCR analysis further validated this effect, revealing a consistent and statistically significant decrease in HIV-1 DNA copy number across biological replicates, underscoring the robustness and reproducibility of our system. Importantly, we also demonstrated that CRISPR-mediated disruption of the provirus effectively blocks HIV-1 reactivation. Treatment with SAHA and TNFα, which typically induce viral reactivation, failed to elicit GFP expression in LNP-treated J-Lat 10.6 and 2D10 cell lines. Confocal microscopy and flow cytometry confirmed a marked reduction in reactivation, suggesting that the proviral genome was rendered transcriptionally inactive or totally eliminated. This functional blockade of HIV-1 reactivation or rebound represents a critical step toward achieving a sterilizing or functional cure for HIV-1. It is important to note that the J-Lat 10.6 and 2D10 latency models used in this study contain well-characterized proviral modifications. J-Lat 10.6 harbors an essentially full-length provirus with a nef→GFP substitution and is replication-incompetent but capable of producing virion-like particles upon reactivation [73], whereas 2D10 contains additional partial deletions in Gag/Pol [74]. These differences do not affect the interpretation of proviral excision but are relevant for contextualizing each model's biological properties.

The ability to target resting CD4 T cells—a population typically resistant to transduction and delivery—is a significant advancement in immunotherapy and gene delivery. RNP delivery by electroporation remains a gold standard for T cells but often requires activation/expansion and has viability tradeoffs [107,108]. Cas9‐RNPs packaged into virus‐like Enveloped Delivery Vehicles (EDVs) decorated with antibody fragments enabled gene editing of CD4 and CD8 T cells in a humanized mouse model [[109], [110], [111]]. DR, as designed ankyrin repeat proteins, offer high specificity and affinity for human CD4 in resting T cells, enabling selective engagement of these otherwise elusive cells. Unlike traditional antibodies, DARPins are smaller, more stable, and can be engineered for multivalency or bi-specificity, enhancing their utility in targeted delivery systems [57,[60], [61], [62], [63]]. In this context, our DR-LNP system represents a novel strategy. While LNPs have been widely used for mRNA delivery, most notably in mRNA vaccines, their application in targeting resting CD4 T cells has been limited due to the cells' quiescent nature and low endocytic activity. Although DARPins have previously been used to bind CD4 T cells or retarget viral vectors [59,78], no studies have demonstrated DARPin-mediated delivery of mRNA, RNP, or cDNA into resting CD4 T cells using a non-viral nanoparticle system. Prior work with CD4-specific DARPins has focused on binding, neutralization, or AAV/phage capsid retargeting, without achieving cytosolic delivery of functional cargo to quiescent T cells. Antibody-targeted LNPs have been reported for CD4 T cells in vivo, but these relied on immunoglobulin ligands and did not address delivery to the latent or resting T cell reservoir [112]. By decorating LNPs with DR, our system likely overcomes these barriers through receptor-mediated endocytosis, facilitating efficient mRNA uptake and expression in a cell type that is otherwise refractory to such interventions. To our knowledge, this is the first demonstration of LNP-mediated Cas9 mRNA/dual sgRNA delivery specifically to resting CD4 T cells using a bipolar dual-DARPin-based targeting strategy, enabling functional uptake, expression, and genome modification. This approach not only broadens the scope of mRNA therapeutics but also opens new avenues for targeting latent HIV-1 reservoirs, modulating immune responses, or engineering T cells in situ (like in vivo CAR-T) without the need for ex vivo manipulation. However, translation to primary cells and in vivo systems remains to be established and constitutes ongoing work.

Conclusion

5

Our findings provide proof-of-concept evidence that DARPin-LNP-mediated delivery of spCas9 mRNA and HIV-1-targeting multiplex sgRNAs can effectively excise latent proviral DNA and prevent viral reactivation. This approach represents a meaningful advancement in the development of gene-editing strategies aimed at achieving a functional cure for HIV-1. While both DARPin-functionalized LNPs and DARPin-modified AAVs have shown promise in enhancing cell-specific delivery, our bipolar dual-DARPin-LNP platform may offer potential advantages in specificity, efficiency, scalability, and safety. Future work will include direct head-to-head benchmarking against monovalent, randomly oriented, and antibody-based conjugation strategies to quantitatively define these benefits alongside antibody-decorated LNP comparators to contextualize targeting performance.

Future studies are warranted to evaluate the efficacy of this platform in primary CD4 T cells from PWH, particularly those on ART, as well as any other immune diseases or immunotherapies involving CD4 immune cells. In vivo studies in humanized mouse models will also be essential to assess biodistribution, pharmacokinetics, immune responses, and long-term safety. DR exhibits high affinity for human CD4 but limited cross-reactivity with non-human primate or murine CD4, constraining in-vivo validation models. Future work should also explore: 1) Optimization of LNP formulations for co-delivery of multiple therapeutic agents; 2) Evaluation of potential effects on T-cell activation or viability; 3) Assessing of immunogenicity and toxicity of repeated dosing; 4) Comprehensive genomic specificity and off-target profiling using established approaches (e.g., GUIDE-seq and targeted deep sequencing of predicted off-target loci), alongside computational prediction pipelines with experiments performed in primary CD4 T cells and PBMCs to quantify specificity under relevant cellular contexts; 5) Expanded profiling across PBMC subsets (including CD4 and non-CD4 populations) to quantify cell-type specificity, coupled with CD4 receptor-blocking and ligand-competition assays to validate CD4-dependent targeting (e.g., anti-CD4 monoclonal antibody blockade with isotype controls and soluble ligand competition); 6) Integration with latency-reversing agents or immune modulators to enhance reservoir clearance; and 7) Development of scalable GMP-compliant manufacturing protocols for clinical translation. Taken together, these future studies explicitly address genomic off-target editing and mechanistic specificity—distinct from general toxicity/viability.

CRediT authorship contribution statement

Subhra Mandal: Data curation, Methodology, Validation, Visualization, Writing – original draft, Writing – review & editing. Abdul Rasheed Baloch: Data curation, Investigation, Methodology, Validation, Visualization, Writing – original draft, Writing – review & editing. Xinxu Yuan: Data curation, Methodology, Validation, Writing – review & editing. Jackson Chen: Data curation, Methodology, Validation. A. Sami Saribas: Data curation, Methodology, Validation, Visualization, Writing – review & editing. Yuanjun Zhu: Data curation, Methodology, Validation. Danmeng Zhang: Data curation, Methodology, Validation, Visualization. Dabbu Jaijyan: Conceptualization, Data curation, Methodology, Validation, Writing – review & editing. Jian Xu: Data curation, Methodology. Reafa Hossain: Data curation, Methodology, Validation, Visualization. Ian Sisto: Data curation, Methodology, Validation, Visualization. Hong Wang: Data curation, Methodology, Resources, Supervision. Xiaofeng Yang: Conceptualization, Funding acquisition, Resources, Supervision, Writing – review & editing. Qingsheng Li: Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Resources, Supervision, Writing – original draft, Writing – review & editing. Wenhui Hu: Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Resources, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing.

Declaration of competing interest

A provisional patent application (#63/910,835) was filed on November 4, 2025, via Virginia Commonwealth University with University of Nebraska-Lincoln and the Wistar Institute related to this work.