Near Sequence Homology Does Not Guarantee siRNA Cross-Species Efficacy

Abstract

Small interfering RNAs (siRNAs) represent a novel class of drugs capable of potent and sustained modulation of genes across various tissues. Preclinical development of siRNAs necessitates assessing efficacy and toxicity in animal models. While identifying therapeutic leads with cross-species activity can expedite development, it may compromise efficacy and be infeasible for certain gene targets. Here, we investigate whether deriving species-active siRNAs from potent human-targeting leads—an approach termed mismatch conversion—can yield potent compounds. We systematically altered potent siRNAs targeting human genes associated with diseases—SOD1 (ALS), JAK1 (inflammation), and HTT (HD)—to generate species-matching variants with full complementarity to their target in NHPs, mice, rats, sheep, and dogs. Variants potency and efficacy were measured in corresponding cell lines. We demonstrate that sequence, position, and number of mismatches significantly influence the ability to generate potent species-active compounds via mismatch conversion. Across tested sequences, mismatch conversion strategy ability to identify a species-active lead varied from 0% to 70%. For SOD1, lead compounds identified from species-focus screening in mouse and dog cells were more potent than leads obtained from mismatch conversion. Thus, a focused screening of therapeutic lead and model compounds may represent a more reliable strategy for the clinical advancement of siRNAs.

Introduction

Small interfering RNAs (siRNAs) are an emerging class of therapeutic oligonucleotides that harness the endogenous RNA interference (RNAi) pathway to turn off disease-related genes in a sustained manner. Following cellular internalization, the siRNA's antisense or “guide” strand is incorporated into Argonaute 2 (Ago2) protein to form the RNA-induced silencing complex, which recognizes and degrades complementary mRNA to prevent target protein expression.^1–3 RNAi is highly conserved across species, with minimal changes in Ago2 sequence and structure.^4,5

Whereas the siRNA sequence defines the gene target, the siRNA chemical scaffold (i.e., modification pattern, structure, and conjugate) drives pharmacokinetic/pharmacodynamic properties. Once a siRNA scaffold is optimized for functional, safe delivery to a target tissue, any known gene sequence in that tissue can be modulated by changing the siRNA sequence. This programmability enables rapid discovery and expedites development pipelines for new drug candidates.^6,7 Indeed, following the discovery and optimization of a trivalent N-acetylgalactosamine-conjugated fully stabilized siRNA platform for selective hepatocyte delivery,^8,9 five siRNA drugs were FDA-approved for liver-related diseases, with dozens more in late-stage clinical trials.^7,9

An essential step toward the clinical development of siRNA drugs is evaluating compound efficacy and safety in animal models with tissue characteristics and disease mechanisms resembling those in humans, such as rodents, non-human primates (NHPs), sheep, and dogs. Identifying siRNAs with cross-species reactivity—that is, targeting a site within the mRNA that is homologous across species—for testing in different model organisms accelerated the preclinical development of siRNA drugs in the early years. However, for some genes, using cross-reactive target sites comes at the cost of siRNA potency,⁹ as both the targeting sequence and location within the mRNA influence RNAi activity.¹⁰ Reduced siRNA potency is unacceptable for clinical leads. Alternative strategies are therefore needed to rapidly identify species-active compounds for potent human targeting leads that are not cross-reactive.

When human-targeting siRNA leads are not complementary to the endogenous target site in a model species, preclinical testing could be performed in humanized models. However, such models are not available for many disease/clinical applications. If delivered to non-humanized models, human-targeting leads with mismatches to the animal model target site might be tolerated, but mismatches often result in reduced or abolished RNAi activity. This is because full complementarity between positions 2–17 of the guide strand of the siRNA (which includes the seed region, positions 2–8) and the target mRNA drives target recognition and RNAi-mediated silencing.^11,12 An alternative strategy that might expand targetability against genes across different species is to convert the mismatches in the guide strand of potent human-targeting lead siRNAs to achieve full complementarity to the target site in model species. We call this approach mismatch conversion.

Here, we systematically evaluated the ability of the mismatch conversion strategy to identify model compounds based on potent human leads. We used previously validated lead siRNAs targeting human SOD1 (a gene involved in ALS),¹³ JAK1 (a major regulator of inflammation),¹⁴ and HTT (genetic cause of Huntington’s disease),¹⁵ and aligned the human target site sequence of each with that in NHP, mouse, rat, sheep, and dog to first identify mismatches between them. We then converted the guide strand mismatches to be fully complementary to the model species mRNA target site to generate siRNA variants. The gene silencing efficacy of each variant was tested in the appropriate cell line for each model organism. The target sequence and number of mismatches affected the success of mismatch conversion in generating species-active compounds. When mismatch conversion, a more economical path, was compared side by side with a focused screening strategy for SOD1, we found that leads generated from species-focused screens for mice and dogs were significantly more potent than leads identified from the conversion strategy. This study provides the first systematic evaluation of the mismatch conversion strategy for generating model organism active compounds. While successful in some circumstances, focused screening for species-active compounds may represent a more consistently robust strategy.

Materials and Methods

Cell culture

HeLa (ATCC, #CCL-2), LLC-MK2 (ATCC, #CCL-7), and McA-Rh7777 (ATCC, #CRL-1601) cells were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM) (Cellgro, #10-013CV). N2a and MDCK2 (ATCC #CRL-2936) cells were maintained in Eagle’s Minimum Essential Medium (EMEM) (ATCC #30-2003). Sheep-cultured fibroblasts were maintained in DMEM (Cellgro, #10-013CV) supplemented with MEM nonessential amino acids (Gibco, REF 11140-050) and sodium pyruvate (Gibco, REF 11360-070). All media was supplemented with 9% fetal bovine serum (FBS) (Gibco, #26140), and all cells were grown at 37°C and 5% CO₂. Cells were split every 3 to 7 days and discarded after 15 passages.

Oligonucleotide synthesis

The complete synthesis methodology was previously published by O’Reilly et al. 2023.¹⁶ Briefly, Oligonucleotides were synthesized by phosphoramidite solid-phase synthesis on a Dr Oligo 48 (Biolytic, Fremont, CA) using 2’-Fluoro RNA or 2’-O-Methyl RNA phosphoramidites with standard protecting groups purchased from ChemGenes, Wilmington, MA. Nonconjugated oligonucleotides were synthesized on a 500 Å UnyLinker support (ChemGenes). Cholesterol-conjugated oligonucleotides were synthesized on a 500 Å tetraethylene glycol cholesterol support (ChemGenes). Phosphoramidites were prepared at 0.1 M in anhydrous acetonitrile (ACN), except for 2’-O-methyl-uridine phosphoramidite, which was dissolved in anhydrous ACN containing 15% anhydrous N,N-dimethylformamide. To activate phosphoramidites, 5-(benzylthio)-1H-tetrazole (0.25 M) in anhydrous ACN was used, and the coupling time was 4 min. Capping of unreacted sites was performed using CAP A (20% 1-methyl-1H-imidazole in ACN) and CAP B (30% 2,6-Lutidine and 20% acetic anhydride in ACN). To oxidize the phosphite (P III) center to the phosphate (P V) center, 0.05 M iodine in pyridine-water (9:1, v/v, Apex Industrial Chemicals) was added for 4 min. To sulfurize the phosphite centers, a 0.1 M solution of 3-[(dimethylaminomethylene)amino]−3H-1,2,4-dithiazole-5-thione in pyridine (ChemGenes) was added for 4 min. For detrytilation reactions, 3% trichloroacetic acid in dichloromethane (Apex Industrial Chemicals) was utilized. After synthesis, the columns were washed with a 10% Diethylamine solution in anhydrous ACN. Oligonucleotides were deprotected using methylamine gas for 1 hour, and residual gas was removed in a vacuum desiccator. They were washed and precipitated using ethanol, and ethanol was removed by heating and vacuum. Final oligonucleotides were eluted with water, and their identity was confirmed using LC-MS. Quantification was done by absorbance at 260 nm, and concentrations were calculated using Beer's law. Antisense and sense strands were then combined, heated at 95 °C, and cooled to anneal the strands and form the siRNA. To simplify synthesis, antisense strand oligos begin and end with a Uridine base.

Oligonucleotide delivery to HeLa, LLC-MK2, N2A, McA-Rh7777, sheep cultured fibroblasts, and MDCK2 cells

Cells were plated in the corresponding optimal media, containing 6% FBS (Gibco, #26140) at 8–10,000 cells per well in 96-well tissue culture plates. Cholesterol-conjugated siRNA was diluted in OptiMEM (Gibco, #31985-088) and added to cells, resulting in a 3% FBS final concentration. Cells were incubated for 72 h at 37°C and 5% CO₂.

mRNA quantification by quantigene

mRNA was quantified using the QuantiGene 2.0 Assay (Affymetrix, #QS0011). Seventy-two hours posttreatment, cells were lysed in 250 µL diluted lysis mixture containing proteinase K (Affymetrix, #13228) for 30 min at 55°C. Probe sets and lysate amounts were validated to be in the linear range and were diluted as specified in the manufacturer’s protocol. Appropriate amounts of lysate and diluted probe sets were added to the capture plate and incubated overnight at 55°C according to the manufacturer’s instructions. The following day, amplification and luminescent detection were carried out as specified in the manufacturer’s protocol. The examples of the identification of the linear range and validation of the assays used are shown in Supplementary Figure S1.

Luminescence was detected on either a Veritas Luminometer (Promega, #998–9100) or a Tecan M1000 (Tecan, Morrisville, NC, USA). The specific mRNAs detected are specified in each graph.

Catalog numbers are human HTT (ThermoFisher, #SA-50339), human SOD1 (ThermoFisher, #SA-10232), and human HPRT (ThermoFisher, #SA-10030). Human HPRT (ThermoFisher, SA10030), Rhesus monkey JAK1 (ThermoFisher, sf 4213912), Dog Sod1 (ThermoFisher, Sf4062428), Dog Jak1 (ThermoFisher, sf4062987), NHP HPRT (ThermoFisher, sf 10356), Rat Jak1 (ThermoFisher, sc-3062255), Sheep Htt (Invitrogen, sf10586), NHP HTT (ThermoFisher, sf 10209), Human JAK1 (ThermoFisher, sa50455), Dog Hprt (ThermoFisher, sf-4062885), NHP HPRT (ThermoFisher, sf10356), Nhp HTT (ThermoFisher, sf10209), Mouse Sod1 (ThermoFisher, sb 10053), Sheep Sod1 (ThermoFisher, Sf4261854), Sheep Htt (ThermoFisher, sf10586), Nhp JAK1 (ThermoFisher, sf4214259), Rat Hprt (ThermoFisher, sc 14898), Rat Htt (ThermoFisher, sc 34753), and Mouse Jak1 (ThermoFisher, sb3029714). The housekeeping gene, HPRT, expression was used to normalize toxicity and cell numbers when appropriate.

mRNA detection assay validation

All QuantiGene mRNA-specific probe sets have been validated to ensure linearity and acceptable signal-to-noise range. Supplementary Figure S1 shows representative data for HPRT and SOD1 in different species cell lines.

Statistics

All data were analyzed using GraphPad Prism 9 software (GraphPad Software, Inc). The corresponding figure legends describe sample size and statistical methods used to analyze individual experiments in detail.

Results

Experimental design for evaluation of model species-specific siRNA variants adapted from potent human-targeting lead siRNAs

Using the mismatch conversion strategy, we generated species-specific siRNA variants based on four previously validated lead siRNAs: two targeting different sites in human HTT mRNA, one targeting a site in human JAK1 mRNA, and another targeting a site in human SOD1 mRNA. Species-specific siRNAs targeting SOD1 and HTT were evaluated in NHP, mouse, rat, sheep, and dog cell lines. JAK1 compounds were evaluated in NHP, mouse, rat, and dog cells. Each siRNA was delivered by passive uptake to each species cell line, and siRNA efficacy was determined by measuring target mRNA expression 72 h posttreatment using Quantigene assay. siRNA potency was evaluated using 7-point dose-response curves.

The sequence and chemical modification pattern of all compounds used in efficacy/potency studies is shown in Supplementary Table S1. Briefly, all siRNAs are cholesterol-conjugated and fully chemically modified with either 2′-Fluoro or 2′-O-Methyl at every ribose and with terminal phosphorothioate backbone modifications (Fig. 1A). The cholesterol conjugate facilitates passive uptake into the cell.¹⁷ 2′-Fluoro, a synthetic analog of RNA, and 2′-O-Methyl (methyl), a naturally occurring modification, suppress immune activation and enhance nuclease resistance. The phosphorothioate backbone stabilizes the siRNA and facilitates association and uptake into cells.^18–21 The siRNA duplexes have a 21-nucleotide guide strand and a 16-nucleotide passenger strand (3′ overhanging guide strand), resulting in an asymmetrical scaffold that further promotes cellular internalization in vitro and contributes to in vivo delivery.²¹

FIG. 1.

Fully chemically stabilized HTT_10150 siRNA efficiently silences huntingtin mRNA in five species cell lines. (A). The chemical structure and modifications of all siRNAs used in this article. (B). HTT_10150 targeting region is fully homologous across five species. Asterisks represent standard nucleobases that simplify oligo synthesis. (C). Cells were treated by passive uptake of HTT_10150 for 72 h (1.5 µM with 2× serial dilutions), QuantiGene. Data are represented as a percentage of UNT (n = 3, mean ± SD). HeLa in light blue, LLC-MK2 in purple, N2A in orange, McA-Rh7777 in red, sheep fibroblasts in pink, and siRNAs are represented in dark blue. SD, standard deviation; UNT, untreated control.

To confirm efficient transfection and productive delivery into each cell line (Fig. 1), we first evaluated the potency of a previously validated siRNA targeting position 10150 of human HTT.²² HTT_10150 siRNA has full complementarity to the HTT target site across all six species (Fig. 1B) and induced significant silencing of HTT mRNA in all species (Fig. 1C). The maximum observed silencing ranged between 49% and 66%, and IC₅₀ values were in the nM range. This work confirms that the described experimental design allows for efficient and productive delivery of all cell lines with fully chemically modified, cholesterol-conjugated siRNAs.

Systematic evaluation of species-specific siRNA variants targeting SOD1 in different cell lines

The siRNA lead that potently silences human SOD1 (referred to as SOD1_123 human) targets position 123 of SOD1 mRNA. The human SOD1-123 site has a single mismatch against the corresponding site in the NHP transcript (at position 13), four mismatches against the mouse and rat SOD1 transcripts, and five mismatches against the sheep and dog SOD1 transcripts (Fig. 2A). We converted each mismatch to generate fully complementary species-specific siRNA variants. SOD1_123 human and all variants were then delivered to cell lines derived from different species (Fig. 2B, C).

FIG. 2.

Systematic evaluation of the mismatch conversion strategy to identify species-active compounds analogs of SOD1_123 siRNA. (A). SOD1_123 siRNA targeting region shows incomplete homology across species. Mismatches to the human targeting site are shown in red. Converted mismatches in the guide strand of the siRNA variant are shown in capitalized letters. Asterisks represent standard nucleobases that simplify oligo synthesis. siRNAs are represented in dark blue. (B). Study design for testing each siRNA variant in each cell line. (C–H). Cells were treated by passive uptake of siRNAs for 72 h (1.5 µM with 2× serial dilutions), QuantiGene. Data are represented as a percentage of UNT (n = 3, mean ± SD). Target mRNA expression following siRNA treatment in (C). HeLa cells, (D). LLC-MK2 cells, (E). N2A cells, (F). McA-Rh7777 cells, (G). Sheep fibroblasts, and (H). MDCK2 cells. SD, standard deviation; UNT, untreated control.

The only species-specific siRNA that exhibited a similar potency to SOD1_123 human (IC₅₀ ∼148nM) in human HeLa cells was SOD1_123 NHP (IC₅₀ ∼201nM), which has a single G-U wobble base pair with the human mRNA target at position 13 of the guide strand. Similarly, SOD1_123 human, which contains an A-C mismatch at position 13 to the NHP SOD1 target site, exhibited similar potency (IC₅₀ ∼653 nM) to SOD1_123 NHP (IC₅₀ ∼416nM) in NHP cells (Fig. 2D). These results suggest a single mismatch is well tolerated by siRNA in this SOD1 mRNA target site context.

The multiple mismatches between the human SOD1 site and the corresponding sites in mouse, rat, sheep, and dog SOD1 mRNA were not tolerated by siRNA. SOD1_123 mouse, rat, sheep, and dog were inactive in HeLa cells, and SOD1_123 human was inactive in mouse, rat, sheep, and dog cell lines (Fig. 2F, G). SOD1_123 mouse exhibited 73% target silencing and an IC₅₀ of ∼375 nM in mouse cells (Fig. 2E), suggesting that converting the four mismatches (C-U at position 6, C-A at position 8, C-U at position 13, and A-C position 15) to restore complementarity to mouse SOD1 produces an efficacious, potent mouse-active siRNA. SOD1_123 dog exhibited modest efficacy in dog cells (50% silencing), suggesting that converting the five mismatches to restore complementarity to dog SOD1 partially restores siRNA activity (Fig. 2H). By contrast, SOD1_123 rat and SOD1_123 sheep were inactive in rat and sheep cells, suggesting conversion of the four mismatches (against rat SOD1) or the five mismatches (against sheep SOD1) could not restore siRNA activity (Figs. 2F, G). Collectively, these results suggest that the number, position, and identity of the mismatch or local context significantly influence the success of the mismatch conversion strategy.

The potency of species-active siRNAs in their corresponding cells, namely SOD1_123 NHP in NHP cells (IC₅₀ ∼416nM) and SOD1_123 mouse in mouse cells (IC₅₀ ∼375nM), was reduced compared to SOD1_123 human in human cells (IC₅₀ ∼148nM). Thus, in the context of the SOD1_123 targeting site, the mismatch compensation strategy did not identify model species-active compounds with comparable efficacy to the human lead. However, our finding that SOD1_123 human exhibits a similar efficacy to SOD1_123 NHP in NHP cells provides a path for the direct use of the human compound for preclinical studies in the NHPs.

Applying the mismatch conversion strategy to lead siRNA targeting human JAK1 produces species-active compounds

The lead siRNA targeting position 1194 of human JAK1 mRNA (referred to as JAK1_1194 human)²³ is fully complementary to NHP JAK1 mRNA, but possesses two mismatches against mouse and rat JAK1 mRNA (mouse and rat targeting sites are homologous), four mismatches against sheep JAK1 mRNA, and two mismatches against dog JAK1 mRNA (Fig. 3A). As expected, JAK1_1194 human potently silenced NHP JAK1 mRNA (88% silencing, IC₅₀ ∼377 nM) but exhibited reduced silencing efficacy against mouse JAK1 (50% silencing) and rat JAK1 (35% silencing) and was completely inactive against dog JAK1 (Fig. 3B–F). JAK1_1194 mouse/rat exhibited near complete silencing in mouse cells (91% silencing, IC₅₀ ∼170 nM) and rat cells (89% silencing, IC₅₀ ∼302 nM) (Fig. 3D, E), suggesting converting the two mismatches to restore complementarity to mouse and rat JAK1 produces an efficacious, potent mouse/rat-specific siRNA. Converting the two mismatches against dog JAK1 generated a siRNA with partial activity (50% silencing at the top dose) in dog cells (Fig. 3F). Similar to the SOD1 results, the number and position of mismatches profoundly impacted the ability to generate model compounds by mismatch conversion.

FIG. 3.

Systematic evaluation of the mismatch conversion strategy to identify species-active analogs for JAK1_1194 siRNA. (A). JAK1_1194 siRNA targeting region shows incomplete homology across species. Mismatches to the human targeting site are shown in red. Converted mismatches in the guide strand of the siRNA variant are shown in capitalized letters. Asterisks represent standard nucleobases that simplify oligo synthesis. siRNAs are represented in dark blue. (B–F). Cells were treated by passive uptake of siRNAs for 72 h (1.5 µM with 2× serial dilutions), QuantiGene. Data are represented as a percentage of UNT (n = 3, mean ± SD). Target mRNA expression following siRNA treatment in (B). HeLa cells, (C). LLC-MK2 cells, (D). N2A cells, (E). McA-Rh7777 cells, and (F). MDCK2 cells. SD, standard deviation; UNT, untreated control.

Applying the mismatch conversion strategy to lead siRNAs targeting human HTT does not produce species-active compounds

The siRNA leads that potently silence human HTT target position 420 (HTT_420) or position 6579 (HTT_6579) of the HTT mRNA. HTT_420 and HTT_6579 are fully complementary to NHP HTT mRNA but possess mismatches against mouse, rat, sheep, and dog HTT. HTT_420 has the same three mismatches against mouse and rat HTT, three mismatches against sheep HTT, and one mismatch against dog HTT (Fig. 4A). HTT_420 was similarly efficacious and potent in human (73% silencing, IC₅₀ ∼132nM) and NHP (77% silencing, IC₅₀ ∼212nM) cells (Fig. 4B, C), but showed reduced efficacy in mouse and rat cells (Fig. 4D, E), and no activity in sheep and dog cells (Fig. 4F, G). Mismatch conversion failed to produce potent species-specific compounds for this site.

FIG. 4.

Systematic evaluation of the mismatch conversion strategy to identify species-active compound analogs for HTT_420 and HTT_6579 siRNAs. (A). HTT_420 siRNA and (H). HTT_6579 siRNA targeting regions show incomplete homology across species. Mismatches to the human targeting site are shown in red. Converted mismatches in the guide strand of the siRNA are shown in capitalized letters. Asterisks represent standard nucleobases that simplify oligo synthesis. siRNAs are represented in dark blue. (B–G, I–N). Cells were treated by passive uptake of siRNAs for 72 h (1.5 µM with 2× serial dilutions), QuantiGene. Data are represented as a percentage of UNT (n = 3, mean ± SD). (B, I). HeLa in light blue. (C, J). LLC-MK2 in purple. (D, K). N2A in orange. (E, L). McA-Rh7777 in red. (F, M). Sheep fibroblasts in pink, and (G, N). MDCK2 cells in green. SD, standard deviation; UNT, untreated control.

HTT_6579 possesses five mismatches against mouse HTT, four mismatches against rat HTT, five mismatches against sheep HTT, and four mismatches against dog HTT (Fig. 4H). HTT_6579 was similarly efficacious in human (74% silencing, IC₅₀ ∼618 nM) and NHP (71% silencing, IC50 ∼318 nM) cells, but displayed minimal efficacy in the other cell lines (Fig. 4). The mismatch conversion strategy was successful for the mouse HTT target site, generating a potent mouse-active siRNA (75% silencing, IC₅₀ ∼144 nM) (Fig. 4K). The approach failed for rats, sheep, and dog HTT, confirming the success of this strategy is highly sequence and mismatch position-specific (Fig. 4).

Comparing efficacies of mouse-specific siRNAs generated by mismatch conversion versus an independent screen

To determine whether the mismatch conversion strategy can produce compounds with comparable potency to those produced from focused screening, we delivered SOD1_123 mouse (Fig. 2A) or a published mouse siRNA, SOD_287, identified from a published screen²⁴ into mouse cells and performed a 7-point dose-response assay. Although both compounds were active, the potency of SOD1_123 (72% silencing IC50∼903 nM) was almost an order of magnitude lower than that of SOD_287 (90% silencing, IC50∼97 nM) conversion (Fig. 5A).

FIG. 5.

Species-specific screening identifies highly potent siRNA Leads. (A). Head-to-head potency comparison of mouse-specific siRNA targeting SOD1 derived from mismatch conversion strategy (left) versus extensive screening (right). (B). Primary screen of siRNAs designed to target dog SOD1 mRNA in MDCK2 (dog) cells. (C). Evaluation of potency of different leads identified from the primary screen in B. Cells were treated by passive uptake of siRNAs for 72 h (1.5 µM with 2× serial dilutions), QuantiGene; Data are represented as a percentage of UNT (n = 3, mean ± SD). SD, standard deviation; UNT, untreated control.

A focused screen in dog cells identified several potent compounds against dog SOD1

The mismatch conversion strategy failed to identify potent dog-active SOD1 compounds (Fig. 2). We, therefore, synthesized a library of 96 fully chemically stabilized cholesterol-conjugated siRNAs targeting various regions of the canine SOD1 gene. The siRNAs were designed using an efficacy prediction algorithm that scores sequences based on their specificity, seed complementarity, and G: C content.²⁵ Compounds were delivered to MDCK2 (dog) cells, and SOD1 mRNA expression was evaluated 72 h posttreatment. This screen identified five efficacious siRNA leads (Fig. 5B). Top leads, evaluated in a 7-point dose response (Fig. 5C), were potent, exhibiting IC₅₀ values ranging from 213 to 470 nM.

The focused screen in dog cells identified species-active lead siRNAs as potent as the human lead ortholog (IC₅₀ ∼148nM), confirming that focused screening is a robust strategy for identifying potent model siRNA compounds.

Discussion

Therapeutic siRNAs are progressing in clinical development, with six drugs currently FDA-approved. Initially, the siRNA lead design favored targeting sites with homology across species. However, this strategy compromises efficacy. Thus, the majority of current siRNA leads for clinical advancement lack perfect homology with animal models.²⁶ Two alternative strategies can be considered to identify model compounds for animal studies: converting existing mismatches in a human lead or performing a parallel focused screen for model compounds. While the conversion strategy is more economical, the species-focus screen may represent a more universal approach. This study explores both strategies and highlights the complexity of finding species-active siRNA variants.

We observed that the sequence and number of mismatches greatly influence the ability to identify functional species-active siRNAs based on the human lead. The data are highly specific to sequence and position, making it challenging to identify general trends. In contrast, focused screening offers a high probability of identifying potent model compounds and thus may represent a better and more consistent approach to facilitate rapid clinical translation. Interestingly, the conversion strategy was successful in finding active mouse compounds for three out of four siRNAs. However, the focused screening strategy requires a significant additional investment of resources and time. In some cases, particularly when the number of mismatches is low (one or two), it might be worthwhile to evaluate if the compounds may be active on their own or if compensating for the mismatch is successful. Indeed, for the SOD1_123 human lead, a single mismatch between humans and NHPs did not significantly affect activity in the NHP cell line. One or two mismatches are often tolerated, even in the seed region, and can sometimes even improve the efficacy of the siRNAs.^27–29

The JAK1_1194 target site data were of particular interest because compensating for two mismatches in mice and rats produced siRNAs that were more potent than the original human lead in human cells. However, results are unpredictable, as compensating for two mismatches in dog JAK1 was completely ineffective.

Initially, we were particularly interested in identifying dog-active siRNAs targeting SOD1. Degenerative myelopathy, a canine disease caused by mutations in the Sod1 gene, could serve as a potential “patient-like” model to explore the therapeutic potential of siRNA targeting Sod1. As conversion strategies did not yield potent SOD1 dog compounds, we conducted a separate screen and identified several functional leads that could be evaluated in a veterinary clinical trial.

In this study, we systematically compare two strategies for identifying species-active model compounds: mismatch conversion and focus species screening. While the mismatch conversion strategy is economically preferred and successful for some compounds, the generated data are not sufficient to make predictions about when this strategy can be valuable, except when sequences are near homologous. The screening approach, while resource-intensive, is guaranteed to generate the model compounds and might be preferred in settings with unlimited resources.

In conclusion, while species-focused screening is the most robust strategy for finding species-active siRNA compounds, the conversion strategy may be used to determine on-target toxicity if only one or two mismatches are present between the species.

Author Disclosure Statement

A.K. is a founder of Atalanta Therapeutics.

Funding Information

The authors would like to thank the National Institutes of Health for supporting this work, including through the S10 OD020012 and R01 NS104022 grants to Anastasia Khvorova and by their support of the Initiative to Maximize Student Development (IMSD) at the University of Massachusetts Chan Medical School, grant T32GM135751 to Brian Lewis, PI, which provided support for I.V.R.F.