To explore whether the increased expression of long interspersed nuclear element 1 (LINE-1; L1) messenger RNA (mRNA) and protein in rheumatoid arthritis synovial fibroblasts (RASFs) is associated with decreased expression of Trex-1, an exonuclease involved in the metabolization of L1 DNA:RNA hybrids.
Chromatin immunoprecipitation was used to detect L1-related p40 protein (L1-ORF1p) binding sequences in RASFs. Luciferase activity was measured in the synovial fibroblasts following cotransfection of the episomal plasmid with pJM105 expressing L1-ORF1p and pGL3-TS3 carrying the target sequence for L1-ORF1p. This luciferase reporter assay was used to compare the activity between RASFs and osteoarthritis synovial fibroblasts (OASFs) and to assess correlations of luciferase activity with the expression of Trex-1 measured by flow cytometry. The expression of Trex-1 mRNA and protein was also compared using real-time polymerase chain reaction, immunohistochemistry, and Western blot analyses. The role of Trex-1 in the L1-ORF1p–mediated luciferase activity assay was studied using interfering RNAs (iRNA) and a Trex-1 expression vector.
Increased luciferase activity occurred after cotransfection of synovial fibroblasts with pJM105 and pGL3-TS3. L1-ORF1p activity was increased in RASFs as compared with OASFs, and this was correlated inversely with the expression of Trex-1. Levels of Trex-1 mRNA and protein were lower in RASFs than in OASFs. After transfection of the L1 expression plasmid, Trex-1 mRNA levels increased in OASFs, but not in RASFs. The addition of iRNA against Trex-1, however, resulted in an enhancement of L1-ORF1p activity in OASFs to the levels measured in RASFs. Overexpression of Trex-1 inhibited 5-azacytidine–induced expression of p38δ MAPK, a gene carrying the TS3 sequence.
The deficiency of Trex-1 in RASFs allows a longer half-life of gene products encoded by active endogenous L1 retrotransposons. This pathway may play a role in diseases in which the cells exhibit a “spontaneous” aggressive behavior.
Trex-1 is the major 3′→5′ DNA exonuclease in mammalian cells, and mutations in the human trex-1 gene can cause Aicardi-Goutières syndrome (1), a disease that is characterized by perturbed immunity. Such mutations also appear to be involved in systemic lupus erythematosus (SLE) (2), in which the failure to process intermediates of nucleic acid metabolism can result in the activation of uncontrolled autoimmunity. Trex-1 prevents the accumulation of single-stranded DNA (ssDNA) fragments derived from endogenous long interspersed nuclear element 1 (LINE-1; L1) retrotransposons (3), the expression of which has been reported in rheumatoid arthritis synovial fibroblasts (RASFs) (4). The promoter of the L1 retroelement is hypomethylated in RASFs (5), reflecting global DNA hypomethylation. The L1-related p40 protein (L1-ORF1p) has nucleic acid binding properties and favors the transcription of specific genes (6). The L1-encoded p150 protein (L1-ORF2p) has a reverse transcription activity, trans-mobilizing Alu RNA and Pol II transcripts, which yields new autoantigenic, hypomethylated DNA fragments (7).
Since we hypothesized that L1-ORF1p interacts with specific DNA sequences, we developed an assay to characterize its binding. The observation that RNA interference against L1-ORF1p induces morphologic differentiation, reduces proliferation, and decreases the expression of some genes (8, 9) suggests that L1-ORF1p might have features of a transcription factor. ORF1p has been reported to be an RNA binding protein (10), and more recent evidence is in accordance with the hypothesis that L1-ORF1p also binds ssDNA (11, 12).
In the present study, we confirm that human L1-ORF1p binds DNA sequences in the promoter regions of various genes and can activate gene expression, as shown by luciferase reporter assays. Among those genes carrying such binding motifs in their promoters are p38δ MAPK (6) and the c-Met oncogene (13). Polyclonal antibodies directed against human L1-ORF1p were generated, and chromatin immunoprecipitation (ChIP) was performed; the obtained DNA fragments were amplified by polymerase chain reaction (PCR), sequenced, and inserted in front of a luciferase reporter gene. Single-stranded DNA derived from endogenous retroelements, such as L1, was observed to accumulate in Trex-1–deficient cells. One role of Trex-1 is to prevent the accumulation of DNA derived from endogenous retroelements (3, 14). We evaluated whether the expression levels of Trex-1 messenger RNA (mRNA) and protein are down-regulated in RASFs, which are characterized by an increased level of L1 gene products. We also present the possible functional consequences of this Trex-1 deficiency in the context of the L1-ORF1p activity in RASFs.
PATIENTS AND METHODS
Synovial tissue was obtained during synovectomy and arthroplastic surgery from 8 patients with RA and 6 patients with osteoarthritis (OA). Synovial fibroblasts were isolated, cultivated in Dulbecco's modified Eagle's medium (DMEM)/10% fetal calf serum (FCS), and used between passages 4 and 6. The project was approved by the local ethics committee.
A commercially available ChIP Assay kit was used according to the manufacturer's protocol (Upstate Biotechnology). Briefly, sonicated DNA fragments of 106 RASFs were divided into 2 fractions. Rabbit anti-human L1-ORF1p polyclonal antibodies (provided by G. Schumann, Paul-Ehrlich-Institut, Germany) were added to fraction 1, while in fraction 2, antibodies were omitted. After immunoprecipitation, a control PCR was performed using primers designed to amplify the 5′-L1-ORF2 homologous region of the p38δ MAPK promoter (GenBank accession no. Z95152, annealing temperature 57°C), as follows: forward primer AGG-TGG-TAT-GAT-GCC-TCC-AG and reverse primer CTT-GAG-GTC-AGG-AGT-TTG-AG. The newly identified target sequences (TS1–TS4) were analyzed.
The TS3 sequence bound by L1-ORF1p and identified by ChIP and PCR (Table 1), as well as mutants (T-to-A mutation in positions 4, 7, and 13), were inserted into the pGL3-Basic vector (Promega), allowing the identification of sequences that lead to transcriptional activation of the firefly luciferase reporter gene. The plasmids were transfected into OASFs (n = 6 patients) and RASFs (n = 8 patients), and luciferase activity was determined.
Table 1. Luciferase activity following transfection of L1-ORF1p–negative RASFs and OASFs with pGL3 luciferase vectors, in the absence or presence of a low dose of DNA hypomethylator 5-azacytidine (5-AzaC) for 72 hours*
5-AzaC at 0.125 μM
Values are the mean ± SD arbitrary units of luciferase activity in osteoarthritis synovial fibroblasts (OASFs) (n = 6) and rheumatoid arthritis synovial fibroblasts (RASFs) (n = 8). The SV40 control was a pGL3 control vector that included an SV40 promoter. The target sequences (TS1–4) inserted in a pGL3 luciferase vector were as follows: for TS1, 5′-ATT-TTT-TTA-AAA-AAA-ATT-AAA-AAA-TTT-TTT-T-3′; for TS2, 5′-ATT-TTT-TTT-TAA-AAA-AAC-CTT-TTA-AAA-AAT-TT-3′; for TS3, 5′-ATT-TGT-TCT-TTT-TGC-TTA-GTC-TTG-CTT-TCG-CTA-TGC-AGG-CTC-TTT-C-3′; and for TS4, 5′-CTT-AAA-AAA-AAT-TTT-AAA-AAA-AGG-GAA-A-3′. P values were determined by Mann-Whitney U test. NS = not significant.
100 ± 19
99 ± 21
69 ± 21
215 ± 64
84 ± 32
102 ± 33
19 ± 8
187 ± 41
79 ± 37
122 ± 21
In the first experiment, transfected OASFs and RASFs were treated for 72 hours with increasing doses (0–0.5 μM) of the DNA hypomethylator 5-azacytidine (5-AzaC; Sigma), and the luciferase activity was analyzed. In the second experiment, we cotransfected the luciferase reporter plasmid pGL3-TS3 1–15 (in which luciferase expression is under control of the p38δ MAPK promoter) with the L1-ORF1p expression plasmid pJM105 (14). In pJM105, the original retrotransposition-competent L1 element L1.2, which is characterized by a missense mutation in L1-ORF2, was set under transcriptional control of the cytomegalovirus (CMV) promoter of the mammalian episomal expression vector pCEP4 (Invitrogen) (14). Transfection was performed 1 day after cell passage (adjusted to 105 cells/TS45 culture flask), using 1 ml DMEM without FCS, a 1:1 ratio of pGL3 (in μg) to CEP4 (in μg), and a 1:3 ratio of vectors (in μg) to FuGene 6 transfection reagent (in μl). Cells were incubated with the transfection mixture for 4 hours at 37°C. Subsequently, 4 ml DMEM/10% FCS was added. The pGL3 control vector (Promega) containing an SV40 promoter and enhancer sequences was used to verify the transfection efficiency, whereas the empty vectors pGL3-Basic (without insert [mock transfection]) and CEP4 were used as negative controls. Cells were harvested 24 and 48 hours after transfection and counted.
Luciferase activity was measured using the MightyLight RLuc Assay kit (Novagen). The light emission (in relative light units [RLU]/second) was measured for 3 seconds. For the calculation of arbitrary units (AU) of luciferase activity in each sample, the mean background values obtained with pGL3-Basic vector (mock transfection control) and the additional control plasmid pRL-null Renilla luciferase (Promega) were subtracted with values set at 0, whereas the mean values obtained with the pGL3 control were set at 100 AU, and the activity in each sample (in AU) was determined according to the following equation:
For Western blotting, nitrocellulose membranes were incubated with murine anti–Trex-1 polyclonal antibodies (ab67192; Abcam) or with rabbit anti-SAPK4 polyclonal antibodies (ab37936; Abcam) diluted in phosphate buffered saline containing 1% bovine serum albumin, and washed again. Trex-1 and p38δ MAPK were detected using goat anti-mouse IgG–horseradish peroxidase (HRP) or swine anti-rabbit IgG-HRP (Dako) and by enhanced chemiluminescence (Amersham).
Trex-1 was also detected by immunohistochemistry on paraffin-embedded sections of synovial tissue, as described previously (4–6). Rabbit anti–Trex-1 polyclonal antibodies (Abcam), goat anti-rabbit biotinylated antibodies, and streptavidin conjugated to alkaline phosphatase were used for tissue staining, and Dako Fast Red substrate was used to reveal the staining. Rabbit preimmune serum was used as a control for the specificity of primary antibodies.
Trex-1–specific interfering RNAs (iRNA) were designed using an online tool (available on http://www.mirnainfo.com) after selecting a 21-mer–specific sequence. The guider strand sequence of the iRNA was 5′- AUC-UGU-GGA-CCU-UAU-GAU-CTT-3′. Transfection of iRNA against Trex-1 was performed on day 0, together with the transfection of pJM105 and/or the pGL3-derived plasmids, using a mixture of 0.1 μg RNA and 6 μl FuGene 6 in 1 ml DMEM without FCS.
Trex-1 mRNA levels were quantified using the TaqMan gene expression assay for Trex-1 (Hs03055245_s1, forward primer GAC-CAT-CTG-CTG-TCA-CAA-CC, reverse primer CCA-AGG-CTG-GGA-CTA-GTG-TT, probe 5′-FAM-TGC-ACA-CCT-GGC-CAC-AAC-CA-3′-TAMRA) and an ABI Prism 7500 Sequence Detector (Applied Biosystems). The relative gene expression was calculated using the ΔΔCt method, using the 18S ribosomal RNA gene as the housekeeping gene and untreated cells as reference.
L1-ORF1p and Trex-1 proteins were measured by flow cytometry in permeabilized cells, using rabbit anti-human L1-ORF1p polyclonal antibodies or rabbit anti-ATRIP polyclonal antibodies (ab59346; Abcam). Swine anti-rabbit IgG–phycoerythrin (Jackson ImmunoResearch) was used as a secondary antibody. Rabbit IgG (Dako) served as a negative control primary antibody.
Finally, OASFs left untreated or treated with 0.125 μM 5-AzaC were transfected with the pCMV6 Trex-1 expression plasmid (NM_016381.2, transcript variant 1) or the corresponding pCMV6 Entry vector (OriGene/LabForce) as a negative control, using FuGene 6 transfection reagent. After 48 hours, the expression of p38δ MAPK was determined by Western blotting.
Wilcoxon's signed rank test was used for comparing 2 dependent samples, and the Mann-Whitney U test was used for comparison of median values from 2 independent samples. Spearman's rank correlation coefficients were used to show possible relationships between variables. The level of significance was set at P values less than or equal to 0.05.
In order to identify genomic target sequences to which L1-ORF1p can bind, ChIP assays were performed using RASFs. L1-ORF1p–DNA complexes were subjected to immunoprecipitation using anti–L1-ORF1p antibodies. Crosslinking was reversed, and the precipitated DNA was purified, PCR-amplified, and cloned. The L1-ORF1p target sequences obtained by ChIP are shown in Table 1. We inserted these DNA fragments into the pGL3-Basic vector, transfected the luciferase reporter plasmids into L1-ORF1p–negative OASFs (n = 6) and RASFs (n = 8), and added a low dose of DNA hypomethylator 5-AzaC for 72 hours in order to trigger the expression of L1-ORF1p from endogenous genomic L1 elements. In a control experiment, we omitted 5-AzaC. Fibroblasts transfected with pGL3 plasmids in which the luciferase reporter gene is under control of the L1-ORF1p target sequences TS1 and TS3 showed a significant increase in luciferase activity after the addition of 5-AzaC (with TS1, mean ± SD 69 ± 21 AU at baseline versus 215 ± 64 AU with 5-AzaC [P < 0.05]; with TS3, 19 ± 8 AU at baseline versus 187 ± 41 AU with 5-AzaC [P < 0.01]) (Table 1). Activation of L1-ORF1p expression did not have any significant effect on luciferase activity in cells transfected with reporter plasmids in which the luciferase gene is controlled by the TS2 and TS4 sequences.
The activity was, in large part, conserved by the 15-bp motif at the 5′ end of TS3 (TS3 1–15 sequence). Mutation analyses demonstrated the importance of the ATT-TGT sequence for L1-ORF1p binding (results not shown). Thus, we selected the TS3 1–15 sequence for further experiments. TS3 1–15–controlled luciferase activity was induced by 5-AzaC in a dose-dependent manner (Figure 1A). Both luciferase activity and induction of L1-ORF1p expression were more sensitive to 5-AzaC in RASFs as compared with OASFs (Figure 1B). This could be due to the partial hypomethylation of the L1 promoters in RASFs. Addition of the hypomethylating drug to RASF and OASF cultures resulted in an increase in the production of L1-ORF1p in RASFs at a lower 5-AzaC concentration than that in OASFs (Figure 1B).
Approximately 50% of cultured RASFs were observed to express endogenous L1-ORF1p during passages 4–5. In general, this expression decreases with the number of passages, and only 30% of RASFs in culture remained positive at passages 7–8. In cultures transfected with the pGL3-TS3 1–15 luciferase reporter plasmid, we observed an up to 2.4-fold increase in luciferase activity in hypomethylated RASFs after passages 7–8 (mean ± SD at passages 4–5 [n = 5 each], 58 ± 17 AU in L1-low RASFs versus 78 ± 7 AU in L1-high RASFs [P < 0.01]; at passages 7–8, 37 ± 14 AU in L1-low RASFs [n = 7] versus 87 ± 8 AU in L1-high RASFs [n = 3] [P = 0.17]).
To confirm a direct and specific interaction between the TS3 sequence and transiently expressed L1-ORF1p in the cellular environment, we cotransfected the pGL3-TS3 1–15 luciferase reporter plasmid and the episomal mammalian expression plasmid pJM105 (14). In pJM105, a CMV promoter controls expression of the L1.2 element that is tagged with a retrotransposition reporter gene and carries a missense mutation in the RT domain of ORF2. The mutation is rendering the originally active L1 element retrotransposition-incompetent, but it facilitates the expression of functional L1-ORF1p. Compared with the luciferase activity observed after transfection of pGL3-TS3 1–15 alone, cotransfection of pGL3-TS3 1–15 with the L1-ORF1p–expressing pJM105 led to a significant, 3–4-fold increase in luciferase activity within 24 hours (P < 0.01 for 8 RASF samples and 6 OASF samples) (Figure 1C). Unexpectedly, we noted that RASFs showed a higher luciferase activity than OASFs in this experimental system. This could be explained by the absence of a mechanism restricting L1-ORF1p expression in RASFs (Figure 1D). We hypothesized that Trex-1 deficiency in RASFs could be a possible cause for this difference.
We first investigated whether Trex-1 is expressed in RA and OA synovial tissue (Figure 2A). Trex-1 was predominantly expressed in the synovial lining layer and on cells surrounding the vessels. OA synovial tissue—considered to be “normal” in these analyses—showed clear staining in the lining, whereas this labeling was reduced or missing in the majority of RA synovial tissue samples. In particular, Trex-1 expression was significantly reduced in the RA synovial lining (Figure 2B). We also investigated whether this Trex-1 expression pattern could be confirmed in cultures of synovial fibroblasts. The deficiency in Trex-1 in RASFs was confirmed by Western blotting and flow cytometry analyses (Figures 2C and D). Accordingly, real-time PCR showed a 6-fold reduction of Trex-1 mRNA in RASFs (Figure 2D).
While pJM105-transfected OASFs exhibited an increased Trex-1 mRNA production (Figure 3A), the level of Trex-1 mRNA remained low in pJM105-transfected RASFs. In addition, the levels of Trex-1 protein in synovial fibroblasts, as measured by flow cytometry, inversely correlated with the extent of luciferase activity obtained 24 hours after cotransfection of pGL3-TS3 1–15 and pJM105 (Figure 3B).
In OASFs, iRNA-mediated knockdown of Trex-1 expression led to a reduction in Trex-1 mRNA levels by 74–88% (Figure 3C). Trex-1 knockdown resulted in the same level of luciferase activity after cotransfection of pGL3-TS3 1–15 and pJM105 in OASFs as that observed in RASFs (Figure 3D), demonstrating that Trex-1 is a factor influencing L1-ORF1p–mediated activation of luciferase expression.
OASFs do not express detectable levels of endogenous L1-ORF1p or p38δ MAPK, but 5-AzaC can increase the expression of both. We investigated the effect of transiently expressed Trex-1 on the 5-AzaC–induced expression of p38δ MAPK in OASFs. Trex-1 overexpression inhibited 5-AzaC–induced p38δ MAPK expression by 64–79% (Figure 3E).
In the present study, we have demonstrated that L1-ORF1p can bind to specific sequences of the human genomic DNA, and the results suggest that the expression of various genes, in addition to p38δ MAPK, could be affected by L1-ORF1p binding. In preliminary experiments, we observed that the TS1, TS2, and TS3 sequences have the potential to activate luciferase expression under basal conditions. In synovial fibroblast cultures that did not express endogenous L1-ORF1p, we showed that transfection with the TS1 and TS3 sequences led to an increase in luciferase activity after treatment with 5-AzaC, which activated, among other proteins, L1-ORF1p production. In addition, we showed that the luciferase activity assay can differentiate between L1-ORF1p–negative and L1-ORF1p–positive RASF cultures. Cotransfection of the luciferase reporter controlled by the TS3 1–15 sequence with the L1 expression plasmid pJM105 allowed us to confirm the direct interaction between L1-ORF1p and the TS3 1–15 binding sequence in the cellular environment.
In RASFs, the luciferase activity response to the transient expression of L1-ORF1p after contransfection was much more pronounced than in OASFs. This cannot be explained by global or L1 promoter–specific hypomethylation alone, since in this experiment, L1 mRNA was expressed in both RASFs and OASFs from the same episomal retrotransposition reporter plasmid, pJM105. We hypothesized that the differences in the activation of the luciferase reporter between the 2 cell lines could be attributed to a mechanism interfering with the action of the transfected L1 template and/or the L1 mRNA that is expressed from it, either through the action of specific microRNA or a deficiency in Tex-1. We found that Trex-1 expression inversely correlated with luciferase activity after cotransfection with the L1 template. Moreover, the levels of Trex-1 mRNA and protein increased after transfection with the L1 expression plasmid in OASFs, but not in RASFs. We demonstrated that the levels of Trex-1 mRNA and protein in RA synovial tissue and in cultured RASFs are reduced relative to that in OA synovial tissue and OASFs. In addition, inhibition of Trex-1 synthesis by iRNA allowed an increase in L1-ORF1p–mediated luciferase activation in these cells and a longer half-life of L1 mRNA, as observed in RASFs. Finally, overexpression of Trex-1 was shown to interfere with the effect of 5-AzaC and/or L1-ORF1p on the p38δ MAPK promoter.
Thus, the difference between OASFs and RASFs in L1 expression as a response to either a DNA demethylating agent or to the intrinsic DNA hypomethylation is at least partially modulated by the action of Trex-1. Trex-1 limits at least the increased expression of L1, which is associated with DNA hypomethylation. In RASFs, the expression levels of Trex-1 mRNA and protein are reduced relative to those in OASFs, and, in contrast to the mechanism in “normal” cells, Trex-1 is not up-regulated in the presence of L1 mRNA.
Trex-1 has properties that are unusual for an exonuclease, because it has a significant preference for particular DNA sequences, e.g., L1 templates. Trex-1–deficient cells accumulate ∼60-bp ssDNA species in the cytoplasm. Such DNA intermediates are not exclusively generated during DNA replication. DNA viruses and retroviruses are additional sources of ssDNA species that could accumulate if a degrading enzyme is lacking (2, 14). The accumulation of endogenous retroviruses triggers the elevated production of type I interferons, which are highly coexpressed with Toll-like receptor 3 (TLR-3)/TLR-7 in RA synovial tissue. In turn, they enhance TLR-3/TLR-7–mediated cytokine production and also TLR-4–mediated responses. Trex-1 deficiency could also lead to autoreactive T cells and represents a mechanism to prevent autoimmunity caused by ssDNA fragments derived from endogenous retroelements. Another role for Trex-1 has been suggested by its association with the SET complex. This protein complex is involved in granzyme A–mediated cell death, a caspase-independent pathway that involves single-stranded DNA damage. Cells with silenced Trex-1 are relatively resistant to apoptosis (2). Thus, a deficiency in Trex-1 could be an additional factor that allows the generation of a hyperplastic synovial tissue that mediates the progressive destruction of articular cartilage and bone.
In conclusion, the expression of L1 gene products is favored in RASFs by a deficiency in Trex-1 protein. This is not related to a mutation in the trex-1 gene, as in SLE, but to a reduced Trex-1 transcription or a decreased mRNA half-life. Hypomethylated and Trex-1–deficient OASFs showed characteristics that mimicked those of RASFs. This cytokine-independent pathway and its deregulation may play a role in the pathogenesis of RA, as well as in other diseases, in which the cells present a “spontaneous” aggressive behavior.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Neidhart had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Neidhart, Schumann, S. Gay.
Acquisition of data. Neidhart, Karouzakis, R. E. Gay.
Analysis and interpretation of data. Neidhart, Karouzakis, Schumann, R. E. Gay, S. Gay.
ROLE OF THE STUDY SPONSOR
Publication of this article was not contingent upon approval by Novartis.
We thank Dr. Haig H. Kazazian, Jr. (Department of Genetics, University of Pennsylvania School of Medicine, Philadelphia) for providing the pJM105 construct.