Bacterial remnants in chloroplast RNA induce immune responses via mouse TLR13 and human TLR7

Abstract

According to the endosymbiotic theory, chloroplasts originate from cyanobacteria. While bacteria-derived molecules trigger immune responses via pathogen-associated molecular patterns, the immunological impact of chloroplast-derived molecules remains unclear. Here, we demonstrate that chloroplast RNA induces immune responses in mammalian cells. Using RNA sequencing analysis of total RNA extracted from Arabidopsis thaliana chloroplasts and nuclear factor-kappa B (NF-κB) reporter assays, we identify a 15-base chloroplast RNA sequence that activates NF-κB in mouse cells via Toll-like receptor (TLR)13. Additionally, a 33-base sequence containing this 15-base motif is identified that activates human cells via TLR7. This 15-base sequence is homologous to the bacterial 23S ribosomal RNA ligand for TLR13 but exhibits distinct reactivity. We also detect this 15-base sequence in the chloroplast RNA from radish, spinach, and broccoli. These findings highlight the retention of bacterial features in chloroplasts and suggest that dietary chloroplast RNA may modulate mammalian immune responses, potentially mimicking probiotic effects.

A 15-base motif in chloroplast RNA activates NFκB in mouse cells via TLR13 and a 33-base sequence containing this motif activates human cells via TLR7. The conserved motif is present in several plant chloroplasts and may modulate mammalian immunity.

Introduction

The endosymbiotic theory is widely accepted as an explanation for the origin of chloroplasts from a bacterial ancestor^1,2. According to this theory, chloroplasts originated from cyanobacteria that were incorporated into eukaryotic cells over 1.2 billion years ago³. Chloroplasts retain several bacterial features, including circular genomes and ribosomes^4–6. Furthermore, in 2016, Hirano et al.⁷ reported that plastids in the moss Physcomitrella patens possess a peptidoglycan wall containing d-amino acids, resembling bacterial cell walls. However, whether these bacterial traits of chloroplasts can elicit immune responses in mammals remains unknown.

In innate immunity, bacterial RNAs act as pathogen-associated molecular patterns (PAMPs), recognized by pattern recognition receptors (PRRs), triggering defense mechanisms essential for pathogen elimination^8,9. Upon PAMP recognition, PRRs activate transcription factors such as nuclear factor-kappa B (NF-κB) and activator protein-1 (AP-1), leading to the production of inflammatory cytokines such as interleukin (IL)-6 and IL-12p40^10,11. PRRs that recognize pathogen-derived RNA include Toll-like receptors (TLRs) TLR3, TLR7, TLR8, and TLR13, as well as retinoic acid inducible gene-I (RIG-I) and melanoma differentiation-associated protein 5 (MDA5) from the RIG-I-like receptor family¹². Beyond their role in pathogen defense, PAMPs are also crucial for maintaining immune system homeostasis^13–16. For example, double-stranded RNA from lactic acid bacteria, recognized by TLR3 on dendritic cells, induces interferon (IFN)-β production and suppresses colitis¹⁶. Given the structural and functional similarities between chloroplast components and bacterial molecules, we hypothesized that chloroplast-derived components may modulate immune responses similarly to probiotics and the intestinal microbiota.

In this study, we demonstrated the immunomodulatory functions of chloroplast RNA using RNA extracted from aseptically cultured Arabidopsis thaliana chloroplasts and NF-κB and AP-1 reporter assays in both mouse and human models. We identified a specific 15-base sequence in chloroplast RNA that triggered immunological activity in mouse macrophages via TLR13. Additionally, we discovered a 33-base sequence containing this 15-base motif that activated NF-κB in human cells via TLR7. Finally, we discuss these findings in the context of RNA structure, chloroplast evolution, and interactions with predators.

Results

Specific segment of chloroplast RNA of A. thaliana activates NF-κB/AP-1 in mouse macrophages

Sterilized chloroplasts were isolated from A. thaliana, and RNA was extracted for analysis. Bioanalyzer profiling revealed four major chloroplast RNA peaks at 975, 1214, 1455, and 1636 nucleotides (nt), along with additional peaks at 112, 141, and 521 nt in the small-molecule region (Fig. 1a). To evaluate the biological activity of the chloroplast RNA, we tested its effect using RAW-Blue, a RAW264.7-based reporter cell line that enables the assessment of NF-κB and AP-1 activation upon PRR stimulation. Chloroplast RNA significantly activated NF-κB/AP-1, with activity levels significantly higher than those observed for RNA extracted from the mitochondrial and microsomal fractions of A. thaliana (Fig. 1b). We hypothesized that a specific segment of chloroplast RNA drives this activity and aimed to identify the active fragment. Chloroplast RNA was further separated by size using denaturing polyacrylamide gel electrophoresis (Fig. 1c), revealing four distinct bands ranging from 1000 to 2000 nt (Bands 1–4). The bands were excised and RNA was recovered from them. RNA sequencing (RNA-Seq) analysis showed that 65.4–89.6% of total reads mapped to different regions of A. thaliana chloroplast ribosomal RNA (rRNA) [Accession: AP000423.1, Fig. 1d]. As previously reported, chloroplast 23S rRNA undergoes post-maturation processing into three fragments of 0.5, 1.2, and 1.1 kb¹⁷. In our study, Band 3 corresponded to the 1.2-kb fragment, whereas Band 4 mapped to the 1.1-kb fragment. When tested for NF-κB/AP-1 activation in RAW-Blue cells, only RNA from Band 4 induced significant activity (Fig. 1e).

Fig. 1. A specific Arabidopsis thaliana chloroplast RNA fragment induces NF-κB/AP-1 activation in mouse macrophages.

a Electropherogram of RNA extracted from A. thaliana chloroplasts analyzed using a bioanalyzer. The numbers at each peak indicate RNA sizes in nucleotides (nt). b NF-κB/AP-1-inducible SEAP activation in RAW-Blue cells, measured based on the absorbance at 620 nm using QUANTI-Blue Solution mixed with the culture medium from RAW-Blue cells stimulated with RNA from each fraction for 16 h. c Gel image of A. thaliana chloroplast RNA separated via denaturing polyacrylamide gel electrophoresis. d Percentages of mapped sequences on each rRNA genome, determined via RNA-Seq analysis of RNA recovered from each band. The numbers in the pie charts correspond to bands labeled no. 1 to no. 4 in c. e NF-κB/AP-1-inducible SEAP activation in RAW-Blue cells measured at 620 nm absorbance in the culture medium from RAW-Blue cells stimulated with RNA recovered from each band for 16 h. Data are presented as mean ± SD from n = 3 independent experiments. Statistical analysis was performed via an ordinary one-way ANOVA with Tukey’s honestly significant difference test. * indicates p < 0.0001. NF-κB nuclear factor-kappa B, AP-1 activator protein-1, SEAP secreted embryonic alkaline phosphatase, ANOVA analysis of variance, SD standard deviations.

Specific sequence in chloroplast rRNA induces activity via TLR13

To determine whether the observed activity was sequence-dependent, we synthesized RNA corresponding to sequences from Bands 3 and 4 using in vitro transcription (IVT). IVT-derived RNA from Band 4 induced NF-κB/AP-1 activation at levels comparable to those of the native Band 4 RNA (Fig. 2a). Systematic deletion analysis identified a 57-nt segment (4_1_2_1_1) within Band 4 that specifically activated NF-κB/AP-1 (Fig. 2b, c). This sequence maps to a region of chloroplast 23S rRNA homologous to the TLR13 ligand sequence found in bacterial 23S rRNA¹⁸. Alignment of A. thaliana chloroplast 23S rRNA with Escherichia coli (Accession: CP026085.1) and Lactobacillus salivarius (Accession: CP000233.1) revealed that the 57-nt sequence differed from the bacterial TLR13 ligand sequence in two nucleotides (Fig. 2d). We then examined the activity of the remainder of the 57-nt segment (4_1_2_1_1) [minus Chloroplast_15b]. We synthesized RNA oligonucleotides using the sequences named “4_1_2_1_1_5ʹ fragment (5′ fragment)” and “4_1_2_1_1_3ʹ fragment (3′ fragment)” as listed in Supplementary Table 1 and assessed their activity using the NF-κB/AP-1 reporter assay with RAW-Blue cells. Neither of the RNA sequences exhibited activity (Supplementary Fig. 1).

Fig. 2. Detailed identification of the chloroplast RNA fragment that activates NF-κB/AP-1.

a NF-κB/AP-1-inducible SEAP activation in RAW-Blue cells measured at 620 nm absorbance in the culture medium of cells stimulated with RNA recovered from Bands 3 and 4 or the corresponding RNA fragments synthesized via in vitro transcription (IVT) for 16 h. Individual data points from two independent experiments (n = 3 per experiment) are presented as dots. b Schematic representation of the strategy used to identify the active fragment in Band 4. Each RNA fragment was synthesized via IVT. c NF-κB/AP-1-inducible SEAP activation in RAW-Blue cells stimulated with IVT-synthesized RNA fragments [as shown in b] for 16 h. d Sequence alignment of A. thaliana chloroplast 23S rRNA, Escherichia coli 23S rRNA, and Lactobacillus salivarius 23S rRNA. Specific segments containing the TLR13 ligand sequence are listed in the table. Colored cells indicate sequences common to all three species. e NF-κB/AP-1-inducible SEAP activation in RAW-Blue cells stimulated with the IVT-synthesized RNA fragment 4_1_2_1_1 [as shown in b] or the chemically synthesized RNA fragment Chloroplast_15b [as shown in d] for 16 h. f NF-κB/AP-1-inducible SEAP activation in HEK-Blue-mTLR13 cells stimulated with the chemically synthesized Chloroplast_15b RNA fragment. g TLR13 expression level in HEK-Blue-mTLR13 cells with TLR13 knocked down by siRNA. h Changes in NF-κB/AP-1-inducible SEAP activation in HEK-Blue-mTLR13 cells stimulated with the chemically synthesized Chloroplast_15b RNA fragment (10 μM) after TLR13 knock down. Data are presented as mean ± SD from n = 3 independent experiments (c, e–h). Statistical analysis was performed via an ordinary one-way ANOVA with Tukey’s honestly significant difference test. * indicates p < 0.0001. NF-κB nuclear factor-kappa B, AP-1 activator protein-1, SEAP secreted embryonic alkaline phosphatase, IVT in vitro transcription, TLR Toll-like receptor, ANOVA analysis of variance, SD standard deviations.

To confirm the functional role of this TLR13 ligand homolog, we performed NF-κB/AP-1 reporter assays using chemically synthesized Chloroplast_15b RNA. This sequence induced NF-κB/AP-1 activation in mouse macrophages (Fig. 2e). In HEK-Blue-mTLR13 cells, Chloroplast_15b activated NF-κB/AP-1, confirming its TLR13-dependent activity (Fig. 2f). The activity was abolished following siRNA-mediated knockdown of TLR13 mRNA in HEK-Blue-mTLR13 cells (Fig. 2g, h, and Supplementary Fig. 2a) and RAW-Blue cells (Supplementary Fig. 2b–d).

Chloroplast_15b activates NF-κB/AP-1 via TLR13 but exhibits less activity than its bacterial homologs

To compare the immunostimulatory activity of Chloroplast_15b with its bacterial counterparts, we examined homologous sequences derived from E. coli (E. coli_15b) and L. salivarius (L. salivarius_15b). NF-κB/AP-1 reporter assays in HEK-Blue-mTLR13 cells revealed dose-dependent activation by all three sequences (Fig. 3a). Notably, Chloroplast_15b elicited significantly lower NF-κB/AP-1 activation than E. coli_15b and L. salivarius_15b (Fig. 3a). To characterize the biological impact of Chloroplast_15b, we conducted RNA-Seq analysis on RAW264.7 macrophages treated with Chloroplast_15b, E. coli_15b, or L. salivarius_15b. A heatmap of the top 1000 most variable genes revealed three distinct expression clusters corresponding to DW (yellow), Chloroplast_15b-treated (green), and L. salivarius_15b- and E. coli_15b-treated cells (blue and pink, respectively) (Fig. 3b). Two major gene clusters were identified. Cluster A genes were highly expressed in DW and Chloroplast_15b-treated cells, and Gene Ontology (GO) analysis indicated enrichment in DNA replication-related pathways (Fig. 3b). In contrast, Cluster B genes were prominently upregulated in cells treated with E. coli_15b- and L. salivarius_15b, with GO terms associated with immune responses and defense mechanisms (Fig. 3b). For Cluster B, analysis of the transcription factors potentially involved using the TRRUST Transcription Factors database in iDEP 2.0 revealed prominent enrichment of Nfkb1-, Rela-, and Rel-target genes (Supplementary Table 2). In contrast, no significant enrichment in transcription factors was observed for the genes in Cluster A.

Fig. 3. Chloroplast_15b RNA activates NF-κB/AP-1 via TLR13 but exhibits activity distinct from that of its bacterial homologs.

a NF-κB/AP-1-inducible SEAP activation in HEK-Blue-mTLR13 cells stimulated with chemically synthesized RNA fragments of Chloroplast_15b, Lactobacillus salivarius15b, and Escherichia coli15b for 16 h. b Heatmap of gene expression levels determined via RNA-Seq analysis of RAW 264.7 cells stimulated with each RNA fragment for 6 h, and the top five pathways identified through Gene Ontology biological process enrichment analysis of Clusters A and B. c Gene expression levels in RAW 264.7 cells stimulated with each RNA fragment for 6 h, determined via real-time PCR. Ccne2 and Gmnn were selected from Cluster A, and IL-1β and TNF-α were selected from Cluster B. Comparison of fold-changes in DEGs induced by Chloroplast_15b vs E. coli_15b (d), Chloroplast_15b vs L. salivarius_15b (e), and L. salivarius_15b vs E. coli (f). The blue lines represent the regression line, with slopes of 0.546 (d) and 0.515 (e). Data are presented as mean ± SD from n = 3 independent experiments. Statistical analysis in (a) and (c) were performed via an ordinary one-way ANOVA with Tukey’s honestly significant difference test. * indicates p < 0.01. NF-κB nuclear factor-kappa B, AP-1 activator protein-1, SEAP secreted embryonic alkaline phosphatase, TLR Toll-like receptor, IL interleukin, TNF-α tumor necrosis factor alpha, ANOVA analysis of variance, DEGs differentially expressed genes, SD standard deviations.

These findings were validated through quantitative real-time PCR. Expression of Cluster A genes (Ccne2 and Gmnn) in Chloroplast_15b-treated cells was approximately 50% that in the control, whereas Cluster B genes (If1b and Tnf) were upregulated relative to those in the control but remained significantly lower than those induced by E. coli_15b- and L. salivarius_15b (Fig. 3c). To clarify the characteristics of the differentially expressed genes (DEGs) induced by Chloroplast_15b, we compared their fold changes with those of the DEGs induced by L. salivarius_ (Fig. 3d) and E. coli_15b (Fig. 3e). Strong correlations were observed between the DEGs induced by Chloroplast_15b and those induced by the two bacterial homologs [r = 0.879 for (Fig. 3d) and r = 0.862 for (Fig. 3e)]. Additionally, the slopes of the regression lines were 0.546 (Fig. 3d) and 0.515 (Fig. 3e), respectively. In contrast, the correlation of DEGs induced by L. salivarius_ and E. coli_15b treatments was r = 0.978, and the slope of the regression line was 0.940 (Fig. 3f).

Mutation at position 2071G mildly reduces NF-κB/AP-1 activation by Chloroplast_15b

To further investigate the relationship between RNA sequence and its TLR13-mediated activity, we synthesized eight Chloroplast_15b variants containing specific point mutations (Fig. 4a). NF-κB/AP-1 reporter assays in HEK-Blue-mTLR13 cells revealed that a mutation at position 2071, converting adenine to guanine (2071A→G), significantly enhanced NF-κB/AP-1 activation. The resulting activity was comparable to that induced by E. coli_15b (Fig. 4b). In contrast, mutations at position 2081 (uracil to cytosine; 2081U→C) and position 2082 (adenine to guanine; 2082A→G)—corresponding to the 13th and 14th nucleotides of the sequence, respectively—did not significantly alter NF-κB/AP-1 activation (Fig. 4b).

Fig. 4. Identification of the nucleotides responsible for modulating NF-κB/AP-1 activation intensity via TLR13.

a Mutant sequences of Chloroplast_15b were designed as shown in the table. Mutated nucleotides are highlighted in red. b NF-κB/AP-1-dependent SEAP activation in HEK-Blue-mTLR13 cells stimulated with each RNA mutant described in a for 16 h. Data are presented as mean ± SD from n = 3 independent experiments. NF-κB nuclear factor-kappa B, AP-1 activator protein-1, SEAP secreted embryonic alkaline phosphatase, TLR Toll-like receptor, SD standard deviations.

Chloroplast_15b sequence is found in the mouse genome, unlike E. coli_15b and L. salivarius_15b

To explore the evolutionary implications of the milder activity of Chloroplast_15b, we searched for sequences in the mouse genome (GRCm39/mm39). Chloroplast_15b was identified at 16 sites on the plus strand, including within protein-coding genes such as Col14a1, Dock4, and Ppp1r14a (Supplementary Tables 3 and 4). In contrast, sequences homologous to E. coli_15b and L. salivarius_15b were not detected in any protein-coding genes, although they were found in other genomic regions (Supplementary Table 3). Interestingly, sequences with mild activity, including Chloroplast_15b, were significantly more frequent in the mouse genome than highly active sequences (Supplementary Fig. 3). To evaluate whether the total of 32 (16/16, plus/minus strands) genomic hits for Chloroplast_15b was unusually high, we conducted a random search of 1000 different 15b sequences within the mouse genome. Approximately 39.4% were absent from the mouse genome, while 15.8% were found at a single site (Supplementary Fig. 4). In contrast, sequences with more than 30 hits, such as Chloroplast_15b, accounted for only 1.8% (Supplementary Fig. 4).

To determine whether Chloroplast_15b is consistently present while E. coli_15b and L. salivarius_15b are rarely detected in other mammals; we analyzed the prevalence of these 15-base sequences across 53 mammalian genomes, including human, using GGGenome (https://gggenome.dbcls.jp/en/) [Fig. 5].

Fig. 5. Genome-wide occurrences of three immunostimulatory RNA sequences in mammals.

Exact matches of three 15-nt sequences (Chloroplast_15b, E. coli_15b, and L. salivarius_15b) were searched across 53 mammalian genomes using GGGenome (https://gggenome.dbcls.jp/en/). Each dot represents one mammalian genome. Paired Wilcoxon signed-rank tests with Holm correction revealed significantly more genomic matches for Chloroplast_15b compared with L. salivarius_15b and E. coli_15b (both p = 2.23 × 10⁻¹⁰).

In total, 31 Chloroplast_15b sequences are also found in the human genome, with 12/19 (plus/minus strands). In contrast, E. coli_15b and L. salivarius_15b are rarely detected, with 0/0 and 2/0, respectively (plus/minus strands). Pairwise comparisons using Wilcoxon signed-rank tests with Holm correction revealed that Chloroplast_15b had significantly more genome matches than L. salivarius_15b (p = 2.23 × 10⁻¹⁰) and E. coli_15b (p = 2.23 × 10⁻¹⁰).

Specific 33b segment in chloroplast RNA induces NF-κB activation via human TLR7

Given that TLR13 is not expressed in humans, we investigated whether a homologous sequence from chloroplast 23S rRNA could activate TLR7, similar to how a sequence in Staphylococcus aureus (SaIIId5) activates TLR7 in mouse plasmacytoid dendritic cells to induce IFN-α¹⁹. We synthesized a homologous sequence of the SaIIId5 region from chloroplast 23S rRNA, referred to as Chloroplast_33b, and tested its ability to activate human TLR7 in 293XL/hTLR7 cells (Fig. 6a). Our results showed that Chloroplast_33b induced IL-8 expression in these cells (Fig. 6b). In contrast, neither Chloroplast_15b nor the 33b sequences derived from the inactive fragment of chloroplast RNA (33b-A and 33b-B from Band 3 in Fig. 1, and Supplementary Table 1) were able to activate TLR7 (Fig. 6a, b). Pre-incubation with the TLR7/TLR8 inhibitor enpatoran completely abolished IL-8 induction (Supplementary Fig. 5). siRNA-mediated knockdown of TLR7 mRNA in 293XL/hTLR7 cells abolished the activity (Fig. 6c, d).

Fig. 6. A specific segment in chloroplast RNA, including the Chloroplast _15b sequence, induces NF-κB activation via human TLR7.

a Sequence alignment of Arabidopsis thaliana chloroplast 23S rRNA, Escherichia coli 23S rRNA, and Lactobacillus salivarius 23S rRNA. Specific segments containing the TLR13 ligand sequence were extracted and are listed in the table. Colored cells indicate sequences common to all three species. Blue arrows denote the sequences of chemically synthesized RNAs used in the experiments shown in b and c. b Interleukin 8 (IL-8) expression in 293XL/hTLR7 cells stimulated with chemically synthesized RNAs (10 μM) for 8 h. Data were standardized to TBP expression; 33b-A and 33b-B represent cells stimulated with randomly synthesized 33-base sequences from Band no. 3 in Fig. 1c. c TLR7 expression levels in 293XL/hTLR7 cells with TLR7 knocked down by siRNA. d IL-8 expression in 293XL/hTLR7 cells stimulated with the chemically synthesized Chloroplast_33b RNA fragment (10 μM) after TLR7 knock down. Data are presented as mean ± SD from n = 3 independent experiments. Statistical analysis was performed via an ordinary one-way ANOVA with Tukey’s honestly significant difference test. * indicates p < 0.0001. NF-κB nuclear factor-kappa B, AP-1 activator protein-1, SEAP secreted embryonic alkaline phosphatase, TLR Toll-like receptor, SD standard deviations.

Chloroplast_15b sequence is highly conserved among green plants and cyanobacteria

To assess its conservation across diverse plant lineages, we analyzed chloroplast 23S rRNA regions corresponding to Chloroplast_15b and Chloroplast_33b in multiple species within each plant order. The results of the nucleotide frequency analyzes are shown in Fig. 7a. Chloroplast_15b was highly conserved within the Brassicales (including A. thaliana) as well as across all other plant groups examined, except for the green algae (Fig. 7a). Although Chloroplast_33b is also highly conserved, our analysis revealed that it exhibits slightly more variability than 15b (Fig. 7a).

Fig. 7. The Chloroplast_15b sequence is derived from cyanobacteria and highly conserved across green plants.

a Nucleotide frequency analysis of sequences corresponding to Chloroplast_33b with a list of vegetables in each order. The numbers in parentheses indicate the number of species (datasets) included in the analysis, where a single representative dataset was used for each species. A summary of the data used for the analysis, along with accession numbers, species names, and taxonomic information, is provided in Supplementary_Data.xlsx. b Phylogenetic tree of bacterial 23S rRNAs with the sequences corresponding to Chloroplast_15b. A single representative dataset was used for each order. Accession numbers used for the phylogenetic analysis are provided in Supplementary_Data.xlsx.

Chloroplast_15b sequence originated from the 23S rRNA of Cyanobacteriota

To investigate the evolutionary origin of Chloroplast_15b, we performed phylogenetic analysis using bacterial 23S rRNA sequences. First, we performed a phylogenetic analysis of 23S rRNA sequences using a dataset of 203 sequences, comprising a single representative species from each bacterial order, including cyanobacteria, mitochondria, and chloroplasts (Supplementary Fig. 5a). Branches corresponding to human and mouse mitochondrial 12S and 16S rRNAs and to A. thaliana mitochondrial 18S rRNA were observed to be phylogenetically distant from bacterial 23S rRNAs. To further elucidate the phylogenetic relationship between mitochondrial rRNA sequences and bacterial 23S rRNAs, we subsequently conducted a phylogenetic analysis with a reduced dataset (Supplementary Fig. 6b). The mitochondrial 18S rRNA of A. thaliana and human and mouse mitochondrial 16S and 12S rRNAs could not be aligned to the Chloroplast_33b region of bacterial 23S rRNAs. Based on these results, only the mitochondrial 26S rRNA of A. thaliana was retained. The remaining mitochondrial sequences were excluded, and the phylogenetic analysis of 23S rRNA 198 sequences was conducted again (Supplementary Fig. 6c). The phylogenetic tree and alignment analysis of the Chloroplast_15b region of the 23S rRNA sequences are shown in Fig. 7b, and the full, unabridged data are available in Supplementary Fig. 7. The Chloroplast_15b sequence “ACAGAAAGACCCUAU” was absent from all bacterial groups except for Cyanobacteriota.

Discussion

In this study, we discovered that chloroplast-derived RNA can activate immune responses in both mouse and human cells through the NF-κB/AP-1 signaling pathway (Figs. 1b and 6b). Bands 1, 3, and 4 collectively contain the entire sequence of the chloroplast 23S rRNA of A. thaliana (Fig. 1c, d) and only Band 4 exhibited activity (Fig. 1e). The Chloroplast_4_1_2_1_1 segment of 57b is essential for activity (Fig. 2c), and neither the 5ʹ nor 3ʹ fragment of Chloroplast_4_1_2_1_1 (4_1_2_1_1_5ʹ and 4_1_2_1_1_3ʹ, respectively) exhibited activity (Supplementary Fig. 1). These results clearly demonstrate that the 15-nt segment in chloroplast RNA, termed Chloroplast_15b (ACAGAAAGACCCUAU) is necessary and sufficient. A key finding was the identification of Chloroplast_15b that closely resembles a previously reported TLR13 ligand sequence (ACGGAAAGACCCCGU) known to activate NF-κB/AP-1 via TLR13 (Fig. 2d)¹⁸. The siRNA knockdown experiments conducted in this study confirmed that Chloroplast_15b also induces an immune response through TLR13 (Fig. 2g, h, and Supplementary Fig. 2a–d). This TLR13 ligand sequence was originally derived from bacterial sources, and the endosymbiotic theory posits that chloroplasts evolved from cyanobacteria, making the identification of this homologous sequence particularly intriguing.

The Chloroplast_15b sequence was highly conserved within Brassicales (including A. thaliana) across most other plant orders (Fig. 7a). This observation suggests that humans may commonly ingest Chloroplast_15b RNA through the consumption of vegetables. Our phylogenetic analysis indicated that the Chloroplast_15b sequence is restricted to Cyanobacteriota (Fig. 7b, and Supplementary Fig. 7). Based on these results, the entire Chloroplast_15b sequence appears to have originated from Cyanobacteriota. Despite its potential biological relevance, to our knowledge, no previous studies have explored the similarity between chloroplast and bacterial RNA sequences in the context of immune activation, thus highlighting a previously unknown role for chloroplasts in immune modulation. Importantly, the sequence retained its immunostimulatory activity, effectively activating immune responses through TLR13, a receptor traditionally associated with bacterial recognition.

A distinctive feature of the Chloroplast_15b sequence is its milder immune activity than that of previously reported bacterial sequences (Fig. 3). Strong positive correlations were observed between the DEGs induced by Chloroplast_15b and those induced by the two bacterial samples (Fig. 3d, e). However, the slopes of the regression lines were less than 1. From these results, we concluded that Chloroplast_15b induces gene expression changes that are highly similar to those triggered by its bacterial homologs, but with weaker intensity. What accounts for this reduced activity? According to Song et al.²⁰, at least four sequence features are required for single-stranded RNA to be recognized by TLR13 and trigger an immune response. These features include: (1) a specific nucleotide at position G2056 (equivalent to A2071 in A. thaliana chloroplast 23S rRNA); (2) A2058, A2060, and G2061 (A2073–G2076 in A. thaliana chloroplast 23S rRNA); (3) C2063 and C2055 (C2078 and C2070 in A. thaliana chloroplast 23S rRNA); and (4) G2057 and C2064 (G2072 and C2079 in A. thaliana chloroplast 23S rRNA). Chloroplast_15b satisfies three of these criteria (2, 3, and 4) but carries a mutation at G2056A (equivalent to G2071A in A. thaliana), which partially retains the ability to induce IL-1β production in RAW264.7 cells¹⁸. Notably, this G2056A mutation occurs naturally in Chloroplast_15b. Our findings revealed that Chloroplast_15b activates NF-κB and induces IL-1β production, albeit at levels weaker than those induced by E. coli_15b and L. salivarius_15b sequences that retain the wild-type G2056 (Fig. 3a, c). Furthermore, IL-1β mRNA expression induced by Chloroplast_15b was lower than that induced by E. coli_15b and L. salivarius_15b (Fig. 3c), consistent with previous findings. Notably, although mitochondria are thought to have evolved from α-proteobacteria, mitochondrial RNA does not trigger as strong an immune response as chloroplast RNA (Fig. 1b). Upon comparison of mitochondrial rRNA sequences with those from chloroplasts, E. coli, and L. salivarius, we identified a similar sequence in the A. thaliana mitochondrial 26S rRNA (“ACGGUAAGACCCCGU”; Accession: NC_037304.1). This mitochondrial sequence meets most essential criteria (1, 3, and 4) for immune activation but harbors a mutation at A2058U (in position 2), which disrupts its ability to induce IL-1β production, consistent with our experimental findings¹⁸ (Fig. 1b).

Another key finding of our study is the discovery of a specific 33b RNA sequence in chloroplasts that stimulates human cells via TLR7 (Fig. 6c, d). In contrast, 33b RNAs randomly generated from inactive fragments of chloroplast RNA (33b_A and 33b_B) failed to activate hTLR7 (Fig. 6b). These results suggest that Chloroplast_33b possesses unique features necessary for TLR7 activation. Previous reports indicate that TLR7 activation requires single-stranded RNAs longer than 3 bases, as well as the presence of free guanosine, conditions that Chloroplast_33b satisfies^21–23. However, the precise mechanism by which Chloroplast_33b selectively activates hTLR7, while other similar sequences do not, remains unclear.

Although chloroplasts share several characteristics with bacteria, they do not replicate or infect host cells in the same manner as pathogens, as most of their genes have been transferred to the host cell nucleus^24–26. This raises an important question: could chloroplast RNA, particularly from ingested food, be detected by the immune system and influence immune function? Chloroplasts do not infect or replicate within the host; therefore, a potential concern is whether chloroplast-derived RNA could reach physiologically meaningful concentrations in mammalian cells. However, RNA sensing by innate immune receptors, such as TLR7 and TLR13, depends on localized and transient RNA accumulation within endosomal compartments, rather than systemic RNA abundance. Non-replicating exogenous RNAs can exert functional effects in vivo. For example, Sato et al.²⁷ demonstrated that intestinally derived bacterial rRNA modulates TLR13-dependent immune homeostasis in the absence of infection. In addition, plant-derived dietary RNAs have been detected in mammalian circulation²⁸.

Consistent with these findings, our results show that chloroplast-derived RNA activates the same innate immune pathways as bacterial homologs but with significantly weaker activity, as evidenced by attenuated transcriptional responses despite strong correlations in gene expression profiles. These results suggest that chloroplast-derived RNA functions as a low-intensity immunomodulatory signal, compatible with physiological dietary exposure rather than pathogenic infection. Nevertheless, future studies should aim to conduct direct in vivo quantification of chloroplast-derived RNA uptake and its local concentrations within immune cell endosomal compartments.

We also explored the evolutionary significance of the Chloroplast_15b sequence by investigating its presence in the mouse genome. Although no E. coli_15b sequence was detected, we identified sequences with mild immune activity, including Chloroplast_15b, at 5–158 loci across the mouse genome (Supplementary Table 3). This feature was conserved in the genomes of other mammals, including humans (Fig. 5). Interestingly, these sequences were also found within the mRNAs of protein-coding genes (Supplementary Table 4). The presence of such sequences in the host genome suggests that an overactive immune response against them could potentially lead to unwarranted inflammation. This highlights an important area for further research: understanding the evolutionary relationship between TLR13 and chloroplast-derived RNA.

Despite our promising results, some limitations of this study must be acknowledged. First, to elucidate the evolutionary relationship, further investigation is needed to determine how the sequence specificity of TLR13 was established and whether chloroplast RNA contributed to this process. Homologs of TLR13 have been identified in organisms such as Xenopus laevis and Ciona intestinalis²⁹. Examining TLR13 expression and sequence recognition specificity in these species could help reconstruct aspects of the receptor’s evolutionary trajectory. Second, future studies should focus on orally administering chloroplast RNA to mice or other experimental animals and comprehensively evaluating the resulting immune responses in vivo. Addressing this limitation will require developing methods for large-scale production of sterile chloroplast RNA, as well as techniques to detect its presence and biological activity in animals. Although research on RNA derived from food remains limited, substantial work has been conducted on RNA-mediated communication in both intercellular and inter-species contexts^30–32. We hope that this study contributes to a deeper understanding of plant–animal communication, an interaction likely refined through millions of years of co-evolution.

In conclusion, to our knowledge, this is the first study to demonstrate that chloroplasts contain RNA sequences capable of activating mammalian cells via mTLR13 and hTLR7. Our findings suggest that chloroplast RNA retains bacterial-like characteristics and may contribute to mediating interactions between plants and mammals, particularly through the immune system. The 15b sequence, commonly found in the chloroplast rRNA of vegetables such as radishes and broccoli, could play a broader role in modulating immune function upon ingestion. Future studies are warranted to investigate whether RNA enclosed within the chloroplast membrane, which shares structural similarities with bacterial cell walls⁷, can influence immune responses when consumed as part of the diet.

Methods

Reagents and cells

RAW-Blue cells (a RAW 264.7 macrophage-derived transfected cell line expressing an NF-κB/AP-1-inducible SEAP [secreted embryonic alkaline phosphatase] reporter gene), HEK-Blue-mTLR13 cells (an HEK293-based co-transfected cell line with the mouse TLR13 gene and an NF-κB-inducible SEAP reporter gene), and 293XL/hTLR7 cells (a HEK293-based transfected cell line expressing human TLR7) were purchased from InvivoGen (San Diego, CA). RAW-Blue cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) [FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan] supplemented with 10% (v/v) heat-inactivated fetal bovine serum (FBS) [Thermo Fisher Scientific, Waltham, MA], 100 μ/mL penicillin (Meiji Seika Pharma, Tokyo, Japan), 100 μg/mL streptomycin (Meiji Seika Pharma), 100 μg/mL Normocin (InvivoGen), and 200 μg/mL Zeocin (InvivoGen). HEK-Blu-mTLR13 cells were maintained in DMEM supplemented with 10% (v/v) heat-inactivated FBS, 100 μ/mL penicillin, 100 μg/mL streptomycin, 100 μg/mL Normocin, 100 μg/mL Zeocin, and 30 µg/mL blasticidin (InvivoGen). 293XL/hTLR7 cells were cultured in DMEM supplemented with 10% (v/v) heat-inactivated FBS, 50 μ/mL penicillin, 50 μg/mL streptomycin, 100 μg/mL Normocin, and 10 µg/mL blasticidin. RAW 264.7 cells were obtained from ATCC (Manassas, VA) and cultured in DMEM supplemented with 10% (v/v) heat-inactivated FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin. RNA oligonucleotides, purified via high-performance liquid chromatography, were purchased from Japan Bio Services (Saitama, Japan), with sequences listed in Supplementary Table 1. We used UltraPure DNase/RNase-free distilled water (Invitrogen) as the solvent for RNA. Primers were obtained from Integrated DNA Technologies (Coralville, IA). The TLR7/TLR8 inhibitor, enpatoran, was purchased from Selleck Biotechnology (Yokohama, Japan)³³.

Plant materials and growth conditions

Arabidopsis thaliana (Columbia-0 (Col-0) ecotype) seeds were obtained from Inplanta Innovations, Inc. (Kanagawa, Japan), surface-sterilized with 0.1% Triton X-100 and 50% hypochlorous acid, and aseptically sown on Murashige and Skoog (MS) medium (Sigma-Aldrich, St. Louis, MO) supplemented with 1% sucrose, 0.05% 2-morpholinoethanesulfonic acid, 0.1% MS Vitamin Solution (Sigma-Aldrich), and 0.5% Agar TC-7 (Ina Food Industry, Nagano, Japan). Plants were grown at 25 °C under continuous light for 18–20 days.

Chloroplast isolation

Chloroplasts were isolated following a previously reported method³⁴ with slight modifications. A 2× chloroplast isolation buffer (CIB) consisting of 0.6 M sorbitol, 10 mM MgCl₂, and 40 mM 4-(2-hydroxyethyl) piperazine-1-ethanesulfonic acid (HEPES; pH 8.0, adjusted with KOH) was prepared, stored at −20 °C, and diluted with an equal volume of distilled water before use. Harvested A. thaliana leaves and stems were washed with CIB and homogenized using a Polytron homogenizer (Kinematika, Eschbach, Germany). The homogenate was filtered through gauze and centrifuged at 1000 × g for 5 min at 4 °C. The obtained pellet was resuspended in CIB, while the supernatant was kept at 4 °C for further fractionation. Chloroplasts were purified using a Percoll gradient (80% and 50%, GE Healthcare Bio-Science AB, Uppsala, Sweden) and centrifuged at 7800 × g for 10 min at 4 °C. The green layer between the 80% and 50% Percoll phases was collected in a tube, washed with HMS buffer (50 mM HEPES, 3 mM MgSO₄, 0.3 M sorbitol, pH 8.0), and stored for further analysis. For mitochondrial isolation, the supernatant was centrifuged at 8000 × g for 20 min at 4 °C. The remaining supernatant, following the mitochondria isolation step, was centrifuged at 100,000 × g for 1 h at 4 °C, and the resultant pellet was used as the microsome fraction.

RNA extraction from chloroplasts

Isolated chloroplasts were homogenized in TRIzol reagent (Invitrogen, Carlsbad, CA) with glass beads (YGBLA-0.3, low alkaline, 0.3 mm, Yasui Kikai Corporation, Osaka, Japan) using a vortex mixer. RNA extraction followed the manufacturer’s instructions. The RNA was treated with DNase I (Nippon Gene, Tokyo, Japan) for 20 min at 37 °C, then purified using the phenol-chloroform method and ethanol precipitation. RNA quality and quantity were assessed using a NanoDrop spectrophotometer (Thermo Fisher Scientific) and an Agilent Bioanalyzer (Agilent, Santa Clara, CA).

Recovery of RNA from denatured polyacrylamide gel

Chloroplast RNA was separated on a denaturing polyacrylamide gel electrophoresis. Gels containing RNA were incubated overnight in gel elution buffer³⁵ at 16 °C. The RNA was then recovered from the buffer using the phenol-chloroform method followed by ethanol precipitation.

Preparation of A. thaliana ribosomal DNA template

RNA extracted from each band of the denaturing polyacrylamide gel was used to synthesize complementary DNA (cDNA) for each rRNA fragment. cDNA synthesis was performed using SuperScript III Reverse Transcriptase (Invitrogen, Carlsbad, CA). The synthesized cDNA was PCR-amplified with PrimeSTAR Max DNA Polymerase (Takara Bio, Shiga, Japan) using the primers listed in Supplementary Table 5. PCR conditions included 35 cycles at 98 °C for 10 s and 68 °C for 90 s. PCR products were separated on a 1% agarose gel, and the desired bands were recovered using the NucleoSpin Gel and PCR Clean-up Kit (Macherey-Nagel, Düren, Germany). The recovered DNA fragments were inserted into the pBluescript KS(+)-BamHI plasmid using the In-Fusion Snap Assembly Master Mix (Takara Bio). Recombinant plasmids were introduced into XL-1 Blue Competent Cells via electroporation and cultured in Luria Broth medium supplemented with carbenicillin at 37 °C for 16 h. Plasmids were extracted using a FastGene Plasmid Mini Kit (Nippon Genetics, Tokyo, Japan), and the sequences of the inserted fragments were verified by using an Applied Biosystems 3500 Genetic Analyzer (Thermo Fisher Scientific) after sample preparation with the SupreDye v3.1 Cycle Sequencing Kit and Performa DTR Gel Filtration Cartridge (EdgeBioSystems, San Jose, CA).

In vitro transcription (IVT)

IVT was performed using the T7 In Vitro Transcription Kit (Takara Bio) according to the manufacturer’s instructions. DNA templates for transcription were prepared via PCR using the primers listed in Supplementary Table 6 and the plasmids described in the previous section. The transcription product was treated with DNase I (Nippon Gene) at 37 °C for 30 min, followed by treatment with calf intestine alkaline phosphatase (Takara Bio) at 37 °C for 3 h, following the method described by Wienert et al.³⁶. The crude RNA was purified using denaturing polyacrylamide gel electrophoresis. The molecular weights of the synthesized RNA were calculated using the Oligo Calc: Oligonucleotide Properties Calculator (http://biotools.nubic.northwestern.edu/OligoCalc.html).

NF-κB and AP-1 reporter assay

RAW-Blue cells were seeded into 96-well plates at a density of 1 × 10⁵ cells/well in an AIM V serum-free medium (Thermo Fisher Scientific) and incubated for 24 h. RNA was diluted in Opti-MEM Reduced Serum Medium (Thermo Fisher Scientific) and transfected into the cells using Lipofectamine 2000 (Thermo Fisher Scientific). After an additional 16 h of incubation, the cell supernatant was collected to determine the SEAP levels. Furthermore, to assess whether the RNA sequences stimulated cells via TLR13, the assay was also conducted using HEK-Blue mTLR13 following the manufacturer’s instructions.

siRNA-mediated knockdown

TLR13 in RAW-Blue cells and HEK-Blue mTLR13 cells were knocked down using stealth siRNAs (Thermo Scientific; control siRNA: 12935200, siRNA_1: MSS219752, and siRNA_2: MSS219753) and Lipofectamine RNAiMAX Transfection Reagent. For TLR7 knockdown in 293XL/hTLR7 cells, stealth siRNAs (Thermo Scientific; control siRNA: 12935200, siRNA_1: HSS121963, and siRNA_2: HSS121964) were used. The knockdown efficiency was confirmed using real-time PCR. The sequences of the primers used are listed in Supplementary Table 7.

Real-time PCR

Real-time PCR was performed using RAW 264.7 and 293XL/hTLR7 cells. Chemically synthesized RNA was added to RAW 264.7 cells using FuGENE 6 (Promega, Madison, WI) for 6 h and to 293XL/hTLR7 cells using the TransIT-X2 Dynamic Delivery System (Mirus Bio, Madison, WI) for 8 h. A TLR7/TLR8 inhibitor, enpatoran, was added to the culture medium 30 min prior to RNA stimulation. Total RNA was extracted using the RNeasy Mini Kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions. cDNA was synthesized with the PrimeScript RT Master Mix (Takara Bio). Real-time PCR was conducted using PowerTrack SYBR Green Master Mix for qPCR (Thermo Fisher Scientific) on a QuantStudio 12K Flex Real-Time PCR System (Thermo Fisher Scientific). The thermal cycling conditions were: initial denaturation at 95 °C for 2 min, followed by 40 cycles of 95 °C for 5 s and 60 °C for 30 s for amplification, and melt curve analysis was performed at the end of the run. The primers used are listed in Supplementary Table 7. Gene expression levels were normalized to those of mouse glyceraldehyde-3-phosphate dehydrogenase or human TATA-box binding protein in each sample, using the comparative Ct method (ΔΔCt).

Phylogenetic analyses

Chloroplast 23S rRNA sequences were retrieved from complete chloroplast genomes in the NCBI Nucleotide database. A single representative RefSeq-registered sequence was selected per species. Cyanobacterial 23S rRNA sequences were extracted from reference genomes with complete status by searching Cyanobacteriota in the NCBI Datasets Genome database. A summary of the data used for base frequency analysis, along with accession numbers, species names, and taxonomic information, is provided in the file Supplementary_Data.xlsx. Sequence logos were generated using the Python package Logomaker (v0.8.6) following a previously reported procedure³⁷.

Bacterial 23S rRNA sequences were obtained from the SILVA LSU database 138.2 Ref NR 99^38,39 and filtered for sequence quality >90, alignment quality >90, and length >2500. A single representative per order was selected for analysis. Accession numbers for the phylogenetic analysis are listed in Supplementary_Data.xlsx. Multiple sequence alignments were generated using MAFFT (v7.310)⁴⁰ with the --auto option. A phylogenetic tree was constructed using FastTree (v2.1.10)⁴¹ with the approximately maximum-likelihood method for nucleotide sequences (GTR+CAT model). The resulting tree was visualized using Interactive Tree of Life (iTOL) [v7.2.2]⁴².

RNA-Seq and analyses of sequencing data

RNA-Seq was performed at Bioengineering Lab (Sagamihara, Japan). Sequencing libraries were prepared using the MGIEasy RNA Directional Library Prep Set and the MGIEasy Circularization Kit (MGI Tech Co., Ltd, Shenzhen, China). DNA Nanoballs (DNBs) were generated using DNBSEQ-G400RS High-Throughput Sequencing Kit (MGI Tech Co., Ltd), and sequencing was performed using the DNBSEQ-G400 platform, producing paired-end 100-base reads.

Cutadapt (version 1.18) was used to trim adapter sequences (Read 1 adaptor: AAGTCGGAGGCCAAGCGGTCTTAGGAAGACAA; Read 2 adaptor: AAGTCGGATCGTAGCCATGTCGTTCTGTGAGCCAAGGAGTTG) and low-quality bases (Quality Value < 20). Short reads (<40 bases) were also filtered out. The processed reads were aligned using HISAT2 (v2.1.0) to either the A. thaliana chloroplast genome (GenBank: AP000423.1, ecotype: Columbia) or the mouse genome (GRCm39, Genome Reference Consortium Mouse Build 39, GCA_000001635.9). Sequence Alignment Map files generated during alignment were converted into Binary Alignment Map files using Samtools (v1.3.1) to facilitate downstream analysis.

Gene expression levels were calculated using StringTie, and the resulting GTF files were compiled into a gene count matrix using prepDE.py (http://ccb.jhu.edu/software/stringtie/dl/prepDE.py)⁴³. Gene count normalization, heatmap construction, and GO enrichment analysis were conducted using iDEP 2.0⁴⁴. A heatmap was constructed using the top 1000 most variable genes. Sequence alignments were performed using MEGA X⁴⁵. Nucleotide sequences were searched within human and mouse genomes using GGGenome (https://gggenome.dbcls.jp/ja/).

Statistics and reproducibility

Data are presented as the mean ± standard deviation and were statistically analyzed using ordinary one-way analysis of variance with Tukey’s honestly significant difference test. All experiments subjected to statistical analysis were performed independently at least three times. The exact n numbers are specified in the figure legends. Statistical analyzes were performed using JMP Pro 17.0.0 (SAS Institute Japan, Tokyo, Japan) or R version 4.4.0, with statistical significance set at p < 0.05. The exact p-values are provided in Supplementary_Data.xlsx.

Author contributions

A.I. and K.K. conceived and designed the study. O.A. and A.I. performed the experiments and analyzed the data. Y.H., Y.I., Hideaki S., Y.Y. and Hiroaki S. contributed to the methodology. A.I. and K.K. wrote the manuscript. All authors reviewed and approved the final version of the manuscript.

Peer review

Peer review information

Communications Biology thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editors: Heejin Yoo and Dario Ummarino.

Funding

This study was supported by JSPS KAKENHI (grant number 21K15281) and JST Moonshot R&D (grant number JPMJMS2021).

Data availability

The raw RNA-Seq datasets analyzed in Figs. 1d and 3b have been deposited in the DDBJ database (https://www.ddbj.nig.ac.jp/index-e.html, BioProject: PRJDB20225 and PRJDB20227). The plasmid sequences generated during the current study are available from the corresponding author on reasonable request. The physical plasmids will be shared with qualified researchers upon request under a Material Transfer Agreement. Uncropped and unedited gel images are provided in Supplementary_Information.pdf as Supplementary Fig. 8. Source data for Figures and Supplementary Figs. are provided in the Supplementary_Information.pdf or Supplementary_Data.xlsx. All other data supporting the findings of this study are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s42003-026-10452-0.