It was recently found that patients with relapsing remitting multiple sclerosis exhibit widespread loss of adenosine-to-inosine (A-to-I) RNA editing, which contributes to the accumulation of immunostimulatory double-stranded Alu RNA in circulating leukocytes and an attendant increase in levels of proinflammatory cytokines (e.g., type I IFNs). A specific Alu RNA (i.e., AluJb RNA) was implicated in activating multiple RNA-sensing pathways and found to be a potent innate immune agonist. Here, we have performed a bioinformatic analysis of A-to-I RNA editing in human melanoma samples and determined that pre-therapy levels of A-to-I RNA editing negatively correlate with survival times, suggesting that an accumulation of endogenous double-stranded Alu RNA might contribute to cancer patient survival. Furthermore, we demonstrated that immunostimulatory Alu RNA can be leveraged pharmacologically for cancer immunotherapy. AluJb RNA was in vitro transcribed and then formulated with endosome-destabilizing polymer nanoparticles to improve intracellular delivery of the RNA and enable activation of RNA-sensing pathways. AluJb RNA/polymer complexes (i.e., Alu-NPs) were engineered to form colloidally stable nanoparticles that exhibited immunostimulatory activity in vitro and in vivo. Finally, the therapeutic potential of Alu-NPs for the treatment of cancer was demonstrated by attenuated tumor growth and prolonged survival in the B16.F10 murine melanoma tumor model. Thus, these data collectively implicate intratumoral Alu RNA as a potentiator of antitumor innate immunity and identify AluJb RNA as a novel nucleic acid immunotherapeutic for cancer.
Loss of A-to-I editing leads to accumulation of unedited Alu RNAs that activate innate immunity via RNA-sensing pattern recognition receptors. When packaged into endosome-releasing polymer nanoparticles, AluJB RNA becomes highly immunostimulatory and can be used pharmacologically to inhibit tumor growth in mouse melanoma models. These findings identify Alu RNAs as a new class of nucleic acid innate immune agonists for cancer immunotherapy.
RNA sensing is a prominent defense mechanism employed by mammalian cells to detect intracellular pathogens and cellular distress (1, 2). The abnormal accumulation of RNA at certain locations in or around a cell is a distinguishing feature of many forms of cell stress (e.g., microbial infection, cellular malfunction, etc.; ref. 3). Accordingly, mammalian cells have evolved to express a class of pattern recognition receptors (PRR) known as RNA-sensing PRRs, which can detect (i.e., bind) and respond to such irregularities by activating innate immunity (4–6). Notably, each RNA-sensing PRR is associated with its own recognition requirements (e.g., RNA sequence, RNA morphology, RNA localization, etc.) and distinctive downstream effects (i.e., gene expression signature), which can also depend on cellular and environmental context (7). Once bound to RNA, the RNA-sensing PRRs trigger intracellular signaling cascades that direct the gene expression of the affected cell and ultimately shape an immune response in a stimulus-specific manner. Thus, PRRs have great potential as therapeutic targets for combating a large variety of diseases, because they can be pharmacologically manipulated to orchestrate host immune responses (8).
Under homeostatic conditions, cellular RNAs tend to be immunologically inert and avoid spurious innate immune activation, because RNA-sensing PRRs can largely discriminate against them based on their subcellular localization, composition, and local concentration. Furthermore, cellular RNAs are naturally processed in a variety of ways to avoid innate immune recognition [e.g., 5′ capping, adenosine-to-inosine (A-to-I) editing, etc.]. However, in certain diseases and pathologic states, mechanisms of RNA processing are impaired, resulting in the activation of RNA-sensing PRRs in the absence of pathogen invasion (i.e., sterile inflammation; ref. 9). Indeed, we have recently discovered that circulating leukocytes of patients with relapsing remitting multiple sclerosis (RRMS) exhibit elevated type-I IFN (IFN-I) levels accompanied by an increased presence of double-stranded Alu RNA derived from endogenous Alu elements in the human genome (10). We also found that, while inactive in an edited single-stranded form, certain Alu RNAs can be highly immunostimulatory in an unedited double-stranded form. We attribute the observed increase in double-stranded Alu RNA to a loss of A-to-I editing, because such posttranscriptional modification can destabilize double-stranded RNA (dsRNA) structures (11). In support of this idea, we found in a subsequent study that Alu RNAs in patients with RRMS exhibit widespread loss of A-to-I editing (12). Furthermore, using RNA sequencing (RNA-seq) analysis, we identified several highly expressed Alu RNAs with elevated levels of A-to-I editing in leukocytes from healthy patients relative to those isolated from patients with RRMS.
Of those endogenous Alu RNAs associated with RRMS, one particular dsRNA of the AluJb class (chr17: 76,418,582 – 76,418,856 on the GrCh37/hg19 assembly) was found to be highly immunostimulatory as determined in a reporter cell screen for both IFN-I and NFκB activity (12). Using various reporter cell knockouts, we also determined that the immunostimulatory capacity of the AluJb RNA is primarily derived from its ability to activate both RIG-I and TLR3 (i.e., RNA-sensing PRRs). Notably, our prior results suggest that MDA5 (i.e., another RNA-sensing PRR) is not involved in the recognition of the AluJb RNA, despite the established involvement of this PRR in the detection of other Alu RNAs. The lack of a response from MDA5 might be attributable to the length of the AluJb RNA (i.e., 274 nt), because longer dsRNA molecules (i.e., >300 bp) are required for robust activation of MDA5 (2).
The cytokine signatures associated with both RIG-I and TLR3 activation typically involve type I IFNs and other proinflammatory cytokines that can exert anticancer effects (13–15). Accordingly, localized inflammation induced by administration of Alu RNA as a pharmacologic agent could be beneficial for the treatment of cancer in the context of promoting the cancer-immunity-cycle [i.e., the process through which the immune system can recognize and kill cancer cells (16)]. Indeed, there are several immunostimulatory RNA therapeutics that activate either the RIG-I or TLR3 pathways now in preclinical and/or clinical development for the treatment of solid tumor cancers (17–20). Notably, we have previously demonstrated that AluJb RNA is more potent in vitro than poly(IC) RNA in terms of both IFN-I and NFκB activity (12), which suggests that the combined activation of RIG-I and TLR3 has potential to enable heightened innate immune activation relative to the discrete activation of either PRR.
Here, we have performed a bioinformatic analysis of A-to-I RNA editing in human melanoma samples and determined that pre-therapy levels of A-to-I RNA editing negatively correlate with survival times within certain cancer patient populations. Loss of A-to-I RNA editing, for example by mutations in ADAR, results in accumulation of Alu dsRNAs that activate a strong innate immune response (3, 4, 9). Therefore, low levels of A-to-I editing in melanoma may also result in accumulation of endogenous double-stranded Alu RNA that might contribute to patient survival. Furthermore, we have “repurposed” an Alu RNA sequence derived from patients with RRMS for the treatment of cancer by engineering a nucleic acid immunotherapeutic, AluJb RNA (Left Arm), which can efficiently target RNA-sensing PRRs and exploit RNA sensing in the context of local cancer immunotherapy (Fig. 1). To enhance intracellular uptake and cytosolic delivery of noncoding, double-stranded, immunostimulatory Alu RNA, AluJb RNA was electrostatically complexed with a polymeric nanocarrier, poly[(DMAEMA)-block-(PAA-co-DMAEMA-co-BMA)] (D-PDB), previously described by our group and others (18, 21–25). Optimized AluJb RNA (Left Arm)/D-PDB nanoparticles (herein referred to as Alu-NPs) were characterized for nanoparticle stability, in vitro transfection efficiency, immunostimulatory activity, and antitumor efficacy. Notably, intratumoral administration of the Alu-NP formulation into murine tumors triggered the expression of proinflammatory cytokines that are known to enhance the tumor infiltration of cancer-killing natural killer cells and T lymphocytes. Finally, therapeutic efficacy of Alu-NP was demonstrated in the B16.F10 murine melanoma tumor model by attenuated tumor growth and prolonged survival. These studies provide proof-of-concept that Alu RNA, here designed to mimic those derived from patients with RRMS, can be leveraged as an innate immune agonist for cancer immunotherapy and establish the Alu-NP platform as a potential new class of RNA therapeutic.
Materials and Methods
Bioinformatic Analysis of A-to-I RNA Editing in Human Melanoma Samples
We obtained whole-genome RNA-seq FASTQ files from the National Center for Biotechnology Information's Gene Expression Omnibus to analyze A-to-I RNA editing in patient melanoma samples. We employed the following workflow to identify A-to-I RNA editing sites from paired FASTQ sequencing files. The main identification tool was a python-based package named the SPRINT toolkit that accepts sequence files and produces text files with the following information for each edit site: (i) genomic location, chromosome, and nucleotides from the p terminus, (ii) type of edit; either A-to-G or T-to-C, strand (+ or −), and (iii) number of edits per site and total number of reads per nucleotide site. Mathematica programs were developed to synthesize data including determination of numbers of editing sites within groups unique or shared by groups, mean numbers of total reads and edits for each editing sites, and mean numbers of total reads and edits per total reads across individual gene regions. This information was tied to an Alu database to annotate each A-to-I editing site: location relative to annotated genes and if intronic, noncoding or RNA gene, 3′ or 5′ untranslated regions, or intergenic and if sites were within Alu or non-Alu elements. For these analyses, we defined three categories of A-to-I editing sites: (i) sites with greater than or equal to 1 edit/per read and greater than or equal to 1 read, “all”, (ii) sites with edit/read ratios greater than or equal to 0.05 and greater than or equal to 5 total reads, and (iii) sites with edit/read ratios greater than or equal to 0.25 and greater than or equal to 10 total reads. To create genome-wide A-to-I–editing indices, we identified all A-to-I–editing sites present in one sample and summed edit/read ratios for all editing sites across the genome.
D-PDB and DB were synthesized via reversible addition-fragmentation chain transfer polymerization as described previously (22). Alu DNA sequences were obtained from the GrCh37 (hg19) assembly. Synthetic Alu DNA with a SP6 promoter incorporated at the 5′ end was obtained from Integrated DNA Technologies (IDT DNA). Alu RNA was reverse transcribed using synthetic Alu DNA templates (IDT DNA) and MEGAscript SP6 (Invitrogen) in overnight reactions at 37°C. Reaction products were treated with Turbo DNase, precipitated with lithium chloride, and purified using the RNeasy MiniElute Cleanup Kit (Qiagen). Alu RNAs were not treated with phosphatases to remove 5′ phosphate groups. To quantitate yields, absorbance was determined at 260 nm. To ensure that the Alu RNAs were of the predicted size, gel electrophoresis was employed.
Nanoparticle formulation was performed as described previously (25). Specifically, lyophilized D-PDB or DB polymers were dissolved in ethanol to a concentration of 50 mg/mL. The polymer stock solution was subsequently diluted in phosphate buffer (pH 7.0, 100 mmol/L) to 10 mg/mL, thereby allowing the polymer chains to self-assemble into micelles. Next, four cycles of centrifugal filtration with Amicon Ultra 0.5 mL Centrifugal Filter Units (Ultracel – 3K, Regenerated Cellulose 3,000 NMWL; MilliporeSigma) were employed to concentrate the polymeric micelles into PBS (pH 7.4; Gibco). An aliquot of the concentrated polymeric micelle solution was taken to determine the resultant polymer concentration. The polymer concentration was determined via UV-vis spectroscopy (Synergy H1 Multi-Mode Microplate Reader; Biotek) based on absorbance at 310 nm using a 96-well plate (REF 655180; Greiner Bio-One). The polymeric micelle solution was diluted in PBS (pH 7.4; Gibco) to 1 mg/mL and then sterile filtered (WHA67801302; MilliporeSigma). The sterile-filtered polymer stock solution was then added to an aqueous solution of the desired nucleic acid at molar ratios corresponding with the desired nitrogen-to-phosphate (N/P) charge ratio. For charge ratio calculations, poly(DMAEMA) at a pH 7.4 was estimated to exhibit 50% protonation. The polymer and nucleic acid mixture was rapidly mixed by pipetting and incubated for 20 minutes at room temperature to allow for complete electrostatic complexation.
Nanoparticle Physical Characterization
Nanoparticle physical characterization was conducted as described previously (25). Dynamic light scattering (DLS; Zetasizer Nano ZS; Malvern Panalytical) was employed to calculate the hydrodynamic sizes of the polymeric micelles and RNA/polymer complexes. Sample was run at pH 7.4 and a polymer concentration of 1 mg/mL. For each 2% agarose gel, 3 g of agarose (16500100; Thermo Fisher Scientific) was dissolved in 150 mL of 1x TAE buffer (REF 46010CM; Corning). The mixture was microwaved until the agarose was fully dissolved and then cast into a gel. The TrackIt 100 bp DNA Ladder (catalog no. 10488058; Thermo Fisher Scientific) was used for all gels unless otherwise indicated. Gel electrophoresis was run at 120 V for 45 minutes. Following electrophoresis, the agarose gels were stained with SYBR Safe dye (S33102; Thermo Fisher Scientific) for 30 minutes in the dark and subsequently imaged (Digital ChemiDoc MP system; Bio-Rad). Transmission electron microscopy (TEM) was conducted on an Osris 200 kV microscope (Tecnai). Briefly samples were dropcast and stained with magnesium tungstate (Nanoprobes) and allowed to dry overnight prior to imaging.
All of the cell lines used in this work were maintained according to supplier specifications. RAW-Dual cells (InvivoGen) were cultured in DMEM (Gibco) supplemented with 4.5 g/L glucose, 2 mmol/L l-glutamine, 10% heat-inactivated FBS (Gibco), 100 U mL−1 penicillin/100 µg mL−1 streptomycin (Gibco), and 100 µg/mL Normocin. To maintain selection pressure of the RAW-Dual cells, 200 µg/mL Zeocin was added every other passage. THP1-Dual cells (InvivoGen) were cultured in RPMI1640 Medium (Gibco) supplemented with 25 mmol/L HEPES, 2 mmol/L l-glutamine, 10% heat-inactivated FBS (Gibco), 100 U mL−1 penicillin/100 µg mL−1 streptomycin (Gibco), and 100 µg/mL Normocin. To maintain selection pressure of the THP1-Dual cells, 10 µg/mL Blasticidin and 100 µg/mL Zeocin were added every other passage. B16.F10 cells (ATCC) and B16.F10 IFN-LUC cells were cultured in DMEM (Gibco) supplemented with 4.5 g/L glucose, 2 mmol/L l-glutamine, 10% heat-inactivated FBS (Gibco), and 100 U mL−1 penicillin/100 µg mL−1 streptomycin (Gibco). To maintain selection pressure of the B16.F10 IFN-LUC cells, 10 µg/mL puromycin was added every passage. All of the cell lines used in this work were cultured in a humidified environment (37°C; 5% CO2) and routinely tested for Mycoplasma contamination.
In Vitro Reporter Cell Assays
The RNA/polymer complexes were screened in 96-well plates (REF 655180; Greiner Bio-One). The wells were seeded with 50,000 reporter cells in 100 µL media. Treatments were diluted to 100 µL in PBS and added to the existing 100 µL media when the wells were approximately 80% confluent. Approximately 24 hours after treatment, cell supernatants were evaluated via the Quanti-Luc (InvivoGen) assay, following manufacturer's instructions. A microplate reader (Synergy H1 Multi-Mode Microplate Reader; Biotek) was used to quantify luminescence and/or absorbance. White, opaque-bottom 96-well plates (REF 655073; Greiner Bio-One) were employed for luminescence measurements. For each sample, signal was calculated from three biological replicates, each with three technical replicates. Signal intensity was normalized by subtracting the average value of a negative control group. For presentation purposes, bell-shaped dose–response curves were truncated at their plateau. Curve fitting analysis (GraphPad Prism software) was employed to estimate EC50 values for the various treatments.
In Vivo Imaging Experiments
An IVIS Lumina III (PerkinElmer) was employed to evaluate the in vivo IFN activity experiment. Approximately 1 × 106 B16.F10 IFN-LUC cells suspended in 100 µL of PBS were inoculated into the rear right flank of 6–8 weeks old C57BL/6 mice (The Jackson Laboratory) via a subcutaneous injection. When tumors reached a volume of approximately 100 mm3, mice were given a single 100 µL intratumoral injection of PBS or Alu-NPs [i.e., AluJb RNA (Left Arm)/D-PDB at an N/P ratio of 4] at a 2 µg RNA dose. Bioluminescence was recorded longitudinally (i.e., 0, 6, 12, and 24 hours). At each predetermined timepoint, the mice were given a dorsal subcutaneous injection of 4.5 mg Pierce d-Luciferin (88293; Thermo Fisher Scientific) solublized in 150 µL PBS, and 15 minutes thereafter, a luminescence image was taken. Mice were shaved and anesthetized via isoflurane gas when necessary.
Quantitative RT-PCR Analysis
Approximately 1 × 106 B16.F10 cells suspended in 100 µL of PBS were inoculated into the rear right flank of 6–8 weeks old C57BL/6 mice (The Jackson Laboratory) via a subcutaneous injection. When tumors reached a volume of approximately 200 mm3, mice were given two 100 µL intratumoral injections administered every 3 days with treatments of either PBS or Alu-NPs [i.e., AluJb RNA (Left Arm)/D-PDB at an N/P ratio of 4] at a 2 µg RNA dose. Mice were euthanized 6 hours after the second intratumoral treatment, and tumors were harvested. A TissueLyser II (Qiagen) was employed to homogenize the tumors, and a RNeasy Plus Mini Kit (Qiagen) was used to isolate tumor RNA.
For the qPCR analysis of proinflammatory gene expression, an iScript cDNA synthesis kit (Bio-Rad) was used according to manufacturer's instructions to reverse transcribe 1 µg of tumor RNA. A CFX Connect Real-time System (Bio-Rad) was employed for qPCR on the resultant cDNA, with the threshold cycle number automatically determined by Bio-Rad CFX manager software V.3.0. The TaqMan gene expression kits (Thermo Fisher Scientific) utilized in this work include mouse Ppib (Mm00478295_m1), mouse Ifnb1 (Mm00439552_s1), mouse Cxcl10 (Mm00445235_ m1), mouse Il6 (Mm00446190_m1), and mouse Tnf (Mm00443258_m1). For each treatment group, at least four biological samples were evaluated with technical duplicates for the each gene. The threshold cycle numbers of each technical duplicate were averaged. Employing the 2−ddCt method of analysis, gene expression was normalized to the house-keeping gene (i.e., Ppib) and subsequently normalized to the negative control (i.e., PBS) treatment values.
In Vivo Tumor Therapy Experiments
Approximately 1 × 106 B16.F10 cells suspended in 100 µL of PBS were inoculated into the rear right flank of 6–8 weeks old C57BL/6 mice (The Jackson Laboratory) via a subcutaneous injection. When tumors reached a volume of approximately 50 mm3, mice were given three 100 µL intratumoral injections administered every 3 days with treatments of either PBS, D-PDB polymer NPs, Edited Alu-NPs, or Alu-NPs [i.e., AluJb RNA (Left Arm)/D-PDB at an N/P ratio of 4] at a 2 µg RNA dose or polymer equivalent. Edited Alu-NPs utilized an inactive form of AluJB RNA (Left Arm). Tumor volume, total murine mass, and murine well-being were recorded every other day. A maximum tumor volume of approximately 1,500 mm3 was the predetermined study endpoint and used as the basis for “survival” analysis. A T-cell depletion study was employed to evaluate the T-cell dependence of treatment efficacy. Mice were inoculated and treated as described above. However, T-cell depletion via intraperitoneal injection of α-CD8 mAb (BE0061; BioXCell) was initiated 24 hours prior to NP treatment. To this end, 100 µg of α-CD8 mAb was injected on days −1, 2, 5, 8, and 11 (Fig. 5F). Tumor volume, total murine mass, and murine well-being were again recorded qod.
The Vanderbilt University Institutional Animal Care and Use Committee (IACUC) reviewed and approved all of the animal experiments performed in this work, and all surgical and experimental procedures were performed in accordance with the regulations and guidelines imposed by the Vanderbilt University IACUC. All mice were maintained under pathogen-free conditions at the animal facilities of Vanderbilt University (Nashville, TN).
For each experiment, statistical significance was determined as indicated in the corresponding figure caption. GraphPad Prism software (Version 7.0c) was employed for all statistical analyses of this work. Plotted values represent experimental means, and error bars represent one SD, with the exception of those in tumor growth plots, which instead represent one SEM for improved visual presentation. ****, P < 0.0001; ***, P < 0.005; **, P < 0.01; *, P < 0.05; ns, not significant.
Data Availability Statement
The data generated in this study are available upon request from the corresponding author.
Analysis of genome-wide A-to-I RNA Editing in Melanoma
We employed the SPRINT software package (26) to measure levels of endogenous A-to-I RNA editing in melanoma tumor samples using whole-genome RNA-seq FASTQ files (27). These analyses scan expressed reads across the genome, identify A-to-G or T-to-C mismatches compared with the reference genome (hg19), filter out known SNPs, report number of mismatches, edits, and number of total reads at each site with a mismatch, and calculate proportion of edits to total reads at each edited site. First, we identified all nucleotide sites across the entire expressed genome with greater than 1 edit and calculated edit/read ratios at each site. Second, to guard against sequencing errors, we identified all editing sites with edit/read ratios greater than 0.05 and total reads greater than 5. We determined both total number of edited sites across the expressed genome and created an A-to-I editing index (AEI) by summing edit/read ratios at each edited site across the expressed genome. We performed this analysis for each melanoma sample with available RNA FASTQ files. These included patient samples both before and after onset of therapy. The study group was comprised of patients with advanced melanoma with the goal of studying tumor responses to either anti-PD-1 therapy, or anti-PD-1 therapy combined with anti-CTLA4. Tumor biopsies were obtained both before (pre-therapy) and 2–4 weeks after (on-therapy) onset of therapy (N = 23); whole-genome RNA-seq and whole-exome sequencing were performed [European Genome Phenome Archive Study ID: EGAS00001004545 (19)].
Negative Correlation Between A-to-I RNA Editing and Time to Death
We next investigated whether there was a statistically significant correlation between the AEI determined in each melanoma sample both prior and after onset of therapy and survival times (Fig. 2). We found a statistically significant negative correlation between survival times and the AEI determined prior to onset of therapy (Fig. 2A). We also found a negative correlation between the AEI determined after onset of therapy and survival times, but this correlation was not statistically significant (Fig. 2B). We next subdivided melanoma samples from the pre-therapy group into those with high (i.e., >50) or low (i.e., <50) AEIs and performed further analysis by employing the Kaplan–Meier estimator (Fig. 2C). We found statistically significant differences in survival times between samples with low or high AEIs, with the low AEI group exhibiting greater survival times and the high AEI group exhibited reduced survival times. The overall median survival time for the low AEI group was also significantly greater than the overall median survival time for the high AEI group (Fig. 2D). The average AEI was not different between the paired pre-therapy and on-therapy groups (pre-therapy group: average AEI = 48 ± 27, on-therapy group: average AEI = 59 ± 35; unpaired t test = 0.25, paired t test = 0.14). Overall, we interpret these results to indicate that initial levels of tumor AEI may contribute to overall survival time in melanoma.
The ADAR (adenosine deaminase acting on RNA) family of genes encodes for ADAR enzymes that convert adenosine to inosine (28), and so we also determined the correlations between the intratumoral pre-therapy AEI and levels of ADAR mRNA and between survival times and levels of ADAR mRNA using the DeSeq package. We found a significant positive correlation between ADAR mRNA levels and intratumoral pre-therapy AEI (Fig. 2E). As with the pre-therapy AEI and survival, we found a negative correlation between ADAR mRNA levels and survival time (Fig. 2F). We interpret these results to indicate that expression levels of intratumoral ADAR contribute to the overall intratumoral AEI and that levels of ADAR may be important factors contributing to overall survival time.
Correlations Between A-to-I RNA Editing Indices and Mutation, Neoantigens, Neopeptides, and Cytolytic Scores
On the basis of previous work by Anagnostou and colleagues investigating correlations between genomic, immunologic, and transcriptomic features and survival in patients with melanoma (27), we evaluated relationships between AEI and several established metrics of tumor immunogenicity. Clinical parameters, including response to therapy, tumor stage and type, and mutation type are summarized (Supplementary Table S1). Major differences between the short and long survival groups were that the short survival groups had more advanced disease, M1C (melanoma has spread to another location that does not involve central nervous system) versus M1A (melanoma has spread only to distant skin and/or soft-tissue sites) and M1B (melanoma has spread to the lung) and greater numbers with progressive disease and failure to respond to therapy. In addition, mutation load, neoantigen load, neopeptide load, and a cytolytic index were determined for each sample. Briefly, mutation load, neoantigen load, and neopeptide loads were determined from whole-exome sequencing data. The cytolytic score was derived from the geometric mean of GZMA and PRF1 mRNA expression levels (19). We compared these additional parameters with the AEI in the pre-therapy group. We did not find a statistically significant correlation between the AEI and mutation load, neoantigen load, neopeptide load, or cytolytic score comparing all samples (Table 1A). We also subdivided samples into low AEI and high AEI and determined whether these two AEI groups exhibited differences in these additional clinical and molecular parameters. Groups with low or high AEI did not exhibit statistically significant differences in mutation load, neoantigen load, neopeptide load, or cytolytic score (Table 1B). Thus, although the AEI is clearly associated with overall survival, the AEI is not associated with these additional metrics of tumor immunogenicity.
|A .||B .|
|Parameter .||Spearman r .||P .||AEI < 50 .||AEI > 50 .||P .|
|Mutation load||0.28||ns||234 ± 266||632 ± 948||ns|
|Neoantigen load||0.17||ns||142 ± 165||363 ± 563||ns|
|Neopeptide load||0.13||ns||379 ± 525||819 ± 1391||ns|
|Cytolytic score||0.29||ns||666 ± 557||303 ± 342||ns|
|A .||B .|
|Parameter .||Spearman r .||P .||AEI < 50 .||AEI > 50 .||P .|
|Mutation load||0.28||ns||234 ± 266||632 ± 948||ns|
|Neoantigen load||0.17||ns||142 ± 165||363 ± 563||ns|
|Neopeptide load||0.13||ns||379 ± 525||819 ± 1391||ns|
|Cytolytic score||0.29||ns||666 ± 557||303 ± 342||ns|
NOTE: (A) Correlation between indicated parameters and the AEI, Spearman r, ns = not significant (# of samples = 22). (B) Average ± S.D. of the indicated parameters were determined in samples with AEI less than 50 and AEI greater than 50, ns = not significant.
Multiple sclerosis is associated with loss of A-to-I editing resulting in accumulation of Alu dsRNAs that stimulate IFN regulatory factor and NFkB transcriptional activity and innate immune activation. In melanoma, as shown here, and several other cancers including head and neck, liver, and breast cancer (29), the AEI is associated with survival rates with high AEI associated with poor survival rates compared with cancers with low AEI. We show here that the intratumoral level of ADAR, the gene that encodes adenosine deaminase specific for dsRNA, the major A-to-I editing enzyme, also positively correlates with the AEI and negatively correlates with survival time in melanoma. High AEIs in cancer may result in diversification of the transcriptome via creation of the equivalent of nonsynonymous mutations arising from differential exon editing, splice variants arising from differential editing in introns, and alterations in mRNA stability arising from differential editing in 3′ untranslated regions (30). In addition, high AEI may result in increased editing of Alu dsRNAs preventing their activation of dsRNA sensors and give rise to loss of innate immune homeostasis. This loss of endogenous Alu dsRNAs produced by high levels of A-to-I editing may contribute to reduced survival rates in individuals with advanced melanoma. To begin to explore this latter hypothesis, we evaluated effects of unedited AluJb dsRNA on innate immune activation, tumor growth, and survival in the mouse B16.F10 melanoma model.
Engineering Alu-NPs to Mimic the Endogenous Expression of Immunostimulatory Alu RNA
The structure-function relationships of the highly immunostimulatory Alu RNA (i.e., AluJb RNA) have recently been reported (12). Because Alu RNAs tend to form two intramolecular double-stranded structures (i.e., left and right arms) separated by an intermediate linker sequence, the importance of each arm in regard to innate immune activation was investigated (12). In our previous studies, each arm was individually in vitro transcribed and then screened with the reporter cells for IFN-I and NFκB activity (12). It was determined that the left arm of the AluJb RNA conserved the same immunostimulatory capacity as the full sequence, while the right arm was completely inactive (12). As a result of this finding—and coupled with the fact that shorter nucleic acid sequences are generally easier to load and deliver with nanocarriers—the AluJb RNA (Left Arm) was utilized herein for studies aimed at investigating the therapeutic application of endogenous Alu RNA.
To explore whether endogenous Alu RNA could be used to stimulate antitumor immunity, immunostimulatory AluJb RNA (Left Arm) was in vitro transcribed, and the endosomolytic polymer, D-PDB was employed to mediate intracellular delivery of the RNA (Fig. 1). While we, and others, have previously leveraged D-PDB for the delivery of nucleic acids, Alu RNAs are unique in their structure and composition, and therefore we first evaluated and optimized the capacity of D-PDB to enhance the activity of AluJb. The synthetic AluJb RNA (Left Arm) was complexed with D-PDB at various N/P charge ratios (i.e., molar amount of protonated amines in the nanoparticle corona/molar amount of phosphates in the nucleic acid) to identify the ratio at which AluJb RNA (Left Arm) is fully loaded onto the polymeric micelles. Gel electrophoresis was performed on the resultant complexes (Fig. 3A), and complete complexation of AluJb RNA (Left Arm) and D-PDB was shown to occur at an N/P charge ratio of 4, which is consistent with previous findings for D-PDB and other nucleic acids (18, 21). At an N/P charge ratio of 4, AluJb RNA (Left Arm)/D-PDB (i.e., Alu-NPs) also demonstrated discrete particle packaging with a uniform particle size distribution centered at approximately 100 nm in diameter (Fig. 3B) with similar particle morphology (Fig. 3C). The lack of particle aggregation also agrees with a previous report (25), which indicates that D-PDB and other similarly sized nucleic acids aggregate only at N/P charge ratios less than 4. Accordingly, an N/P charge ratio of 4 was used for all subsequent experiments involving AluJb RNA (Left Arm)/D-PDB complexes.
The immunostimulatory activity of Alu-NPs was then investigated using in vitro activity assays in reporter cell lines. Alu-NP complexes were administered to IFN-I reporter cells that express a secreted luciferase downstream of IFN-stimulated response elements, and 24 hours after treatment, the relative IFN-I production was measured via luminescence (Fig. 3D and E). In both murine RAW-Dual macrophages and human THP1-Dual monocytes, the Alu-NPs demonstrated potent induction of cellular IFN-I production with EC50 values of 5.3 nmol/L and 8.3 nmol/L, respectively. These results are in stark contrast to the complete lack of immunostimulatory activity demonstrated in vitro by freely administered AluJb RNA (Left Arm) or NPs complexed with edited, inactive AluJB (Left Arm). D-PDB NPs without RNA (i.e., empty NPs) failed to elicit a response in RAW-Dual macrophages and induced only a modest response in THP1-Dual monocytes. Furthermore, while AluJB (Left Arm) complexed with non-endosomolytic DMAEMA-block-BMA (D-B) polymer slightly stimulated type-I IFN production in RAW-Dual macrophages, it failed to activate a response in THP1-Dual monocytes. Thus, D-PDB was necessary for effective intracellular delivery of AluJb RNA, and only the unedited form of AluJb RNA was able to potently activate RNA sensing pathways.
Alu-NPs Stimulate Innate Immunity In Vivo
To determine whether the Alu-NPs conserved their immunostimulatory activity when administered in vivo, they were injected intratumorally twice, spaced 3 days apart, in mice bearing subcutaneous B16.F10 tumors, and 6 hours after the second treatment, tumor RNA was collected for qPCR analysis (Fig. 4A). Gene expression changes were quantified for various proinflammatory cytokines that are known to be associated with RIG-I and TLR3 activation (i.e., Ifnb1, Cxcl10, Il6, and Tnf). Intratumorally administered Alu-NPs elicited significantly increased gene expression of Ifnb1, Cxcl10, and Tnf in the tumor microenvironment (TME). Il6 expression also tended to increase relative to PBS-treated tumors, but in a statistically insignificant manner (P = 0.086).
Subsequently, B16.F10 IFN-LUC tumors, which express luciferase in response to IFN signaling (31), were used to confirm expression of IFN in the TME and to study the kinetics of signaling in response to intratumorally administered Alu-NPs (Fig. 4B and C). The results demonstrate that Alu-NPs can stimulate significant IFN production within the TME, with peak signal detected at 12–24 hours after treatment. Collectively, these data demonstrate that intratumoral delivery of Alu-NPs could stimulate an innate immune response in the TME.
Alu-NPs Exert Antitumor Effects in a CD8+ T Cell–dependent Manner
The therapeutic effect of intratumorally administered Alu-NPs was evaluated in the B16.F10 murine melanoma tumor model (Fig. 5). C57BL/6 mice bearing B16.F10 murine melanoma tumors were treated with either vehicle (PBS) or Alu-NPs for a total of three intratumoral injections, each spaced 3 days apart. Relative to PBS treatment, Alu-NP treatment significantly inhibited tumor growth (Fig. 5A), prolonged survival (Fig. 5C), and was well tolerated by the mice as demonstrated by nonsignificant differences in total body mass during or after treatment (Fig. 5B). Importantly, Alu-NPs delayed tumor growth compared with PBS (vehicle) and D-PDB NPs without RNA (Fig. 5D), and neither NPs without RNA nor NPs complexed with an edited inactive Alu RNA conferred the survival benefit of Au-NPs (Fig 5E). To determine whether efficacy was T-cell dependent, CD8+ T cells were antibody depleted with the initiation of CD8+ T-cell depletion one day prior to the onset of treatment. When Alu-NPs were administered in conjunction with CD8+ T-cell depletion, no inhibition of tumor growth was observed, demonstrating that CD8+ T cells are required for efficacy (Fig. 5F). These data demonstrate that Alu RNA sequences identified in patients with RRMS can be effectively repurposed for cancer immunotherapy in the form of Alu-NPs.
Our analysis has revealed a negative correlation between endogenous intratumoral RNA AEI and melanoma patient survival. We also find a negative correlation between intratumoral ADAR levels, which encodes the major RNA A-to-I editing enzyme, ADAR, and patient survival. These findings are confirmed by comparison of survival times of patients subdivided into high AEI and low AEI groups using Kaplan–Meier survival curves. In contrast, intratumoral AEI does not correlate with other immune parameters including mutation load, neoantigen load, neopeptide load or cytolytic score. The most common A-to-I editing sites are Alu RNAs. Unedited Alu RNAs are potent activators of innate immune responses while edited Alu RNAs are very weak activators of innate immune responses. Increased intratumoral AEI is predicted to result in an increase in levels of edited Alu RNAs compared with unedited Alu RNAs that, in turn, may reduce intratumoral innate immune activation by endogenous immunostimulatory Alu RNAs resulting in decreased survival. To further explore this notion, we tested efficacy of an unedited Alu RNA in the B16.F10 melanoma mouse model.
Double-stranded Alu RNA was successfully “repurposed” as a novel cancer immunotherapy from its role in RRMS-related inflammation. Specifically, immunostimulatory AluJb RNA (Left Arm) was formulated with the endosomolytic polymer, D-PDB to form Alu-NP, which enables the intracellular delivery of RNA and targeted activation of RNA-sensing PRRs. At an N/P charge ratio of 4, Alu-NP complexes exhibited high RNA loading and formed uniform nanoparticles of approximately 100 nm in diameter that could potently activate innate immune signaling pathways as determined with in vitro reporter cell assays. As expected, RNA complexation leads to larger NPs diameter than D-PDB alone, but spherical morphology is conserved. Interestingly, D-PDB alone NPs elicited a type I IFN response in THP-1 Dual monocytes, but not RAW-Dual macrophages, while D-B Alu-NPs treatment resulted in the opposite effect. The cell type–specific changes are likely brought on by activation of different PRRs that are preferentially expressed by RAW-Dual Macrophages or THP1-Dual monocytes (32). Upon intratumoral injection, Alu-NPs induced changes in the expression of key immune-related genes, which altered the cytokine profile of the TMEs, resulting in antitumor effects in the B16.F10 murine melanoma tumor model. The combination of in vitro (Fig. 3) and in vivo (Fig. 5) data supports the conclusion that a carrier delivering active RNA is necessary for efficacy. Notably, free RNA does not stimulate type I IFN production, and polymer alone and Alu-NPs with edited RNA do not provide survival benefit relative to PBS treatment. To the best of our knowledge, this work is the first to demonstrate that immunostimulatory Alu RNAs can be delivered using nanotechnology to stimulate antitumor immune responses.
Regarding future work and applications of Alu-NPs, further characterization of the pharmacokinetic properties of the nanoparticle may be necessary to fully understand the kinetics of signaling and its associated effects on innate immune activation. Moreover, comprehensive cellular and molecular immunophenotyping of the TME in response to Alu-NP remains necessary to establish the major cellular effectors responsible for mediating antitumor responses. However, the lack of antitumor efficacy of Alu-NPs when CD8+ T cells are depleted shows that CD8+ T cells are necessary and provides a starting point for future mechanistic studies. In addition, other nanocarriers (e.g., lipid nanoparticles) exist and have potential to further enhance delivery of Alu RNA, including advanced polymeric carriers recently described by our group (33, 34), and therefore optimization of Alu-NP formulations with improved properties and enhanced therapeutic efficacy will be important in the translation of Alu RNA for cancer immunotherapy. Finally, the immunostimulatory properties of Alu RNA could also be leveraged for other applications; for example, as a vaccine adjuvant, because vaccines necessitate the use of adjuvants that enable the activation of innate immunity to provide long-lasting protection against pathogenic challenge. Collectively, our studies identify Alu RNA, here designed to mimic that derived from patients with RRMS, as a potent innate immune agonist with potential utility as a cancer immunotherapy.
No disclosures were reported.
K.M. Garland: Data curation, investigation, visualization, methodology, writing-original draft, writing-review and editing. A.J. Kwiatkowski: Data curation, investigation, methodology, writing-review and editing. J.T. Tossberg: Resources, data curation, formal analysis, investigation. P.S. Crooke: Formal analysis, investigation, methodology. T.M. Aune: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing. J.T. Wilson: Conceptualization, resources, supervision, writing-original draft, writing-review and editing.
We would like to thank Professor Craig Duvall (Vanderbilt University) for the use of their IVIS imaging instrument and the Vanderbilt Institute of Nanoscale Science and Engineering (VINSE) for the use of their Malvern Zetasizer Nano ZS. The research presented here was supported by grants from the National Science Foundation CBET-1554623 (J.T. Wilson), a Vanderbilt Ingram Cancer Center (VICC) Ambassador Discovery Grant (J.T. Wilson), an American Cancer Society Institutional Research Grant (IRG-58-009-56, to J.T. Wilson), and the Congressionally-Directed Medical Research Program (W81XWH-20-1-0624, to J.T. Wilson) and by grants from the NIH (R21AI144193, to T.M. Aune and R21AI164107-01A1, to T.M. Aune and J.T. Wislon). AJK was supported by T32-CA009592.
Note: Supplementary data for this article are available at Cancer Research Communications Online (https://aacrjournals.org/cancerrescommun/).