Sustained high expression of multiple APOBEC3 cytidine deaminases in systemic lupus erythematosus

APOBEC3 (A3) enzymes are best known for their role as antiviral restriction factors and as mutagens in cancer. Although four of them, A3A, A3B, A3F and A3G, are induced by type-1-interferon (IFN-I), their role in inflammatory conditions is unknown. We thus investigated the expression of A3, and particularly A3A and A3B because of their ability to edit cellular DNA, in Systemic Lupus Erythematosus (SLE), a chronic inflammatory disease characterized by high IFN-α serum levels. In a cohort of 57 SLE patients, A3A and A3B, but also A3C and A3G, were upregulated ~ 10 to 15-fold (> 1000-fold for A3B) compared to healthy controls, particularly in patients with flares and elevated serum IFN-α levels. Hydroxychloroquine, corticosteroids and immunosuppressive treatment did not reverse A3 levels. The A3AΔ3B polymorphism, which potentiates A3A, was detected in 14.9% of patients and in 10% of controls, and was associated with higher A3A mRNA expression. A3A and A3B mRNA levels, but not A3C or A3G, were correlated positively with dsDNA breaks and negatively with lymphopenia. Exposure of SLE PBMCs to IFN-α in culture induced massive and sustained A3A levels by 4 h and led to massive cell death. Furthermore, the rs2853669 A > G polymorphism in the telomerase reverse transcriptase (TERT) promoter, which disrupts an Ets-TCF-binding site and influences certain cancers, was highly prevalent in SLE patients, possibly contributing to lymphopenia. Taken together, these findings suggest that high baseline A3A and A3B levels may contribute to cell frailty, lymphopenia and to the generation of neoantigens in SLE patients. Targeting A3 expression could be a strategy to reverse cell death and the generation of neoantigens.


Results and discussion
Characteristics of the SLE patients. Patients' baseline characteristics are described in Table 1. Mean (± standard deviation) age at sample collection was 32.9 ± 12.5 years. Mean disease duration was 8.8 ± 8. The majority of patients had a clinically and/or serologically active disease. Thirty patients suffered from severe flares, 11 from mild or moderate flares and 16 had no flares. The median (range) SELENA-SLE disease activity index (SLEDAI) score was 6 (0-32). Thirty (57%) patients had a positive Farr assay and 23 (40%) had low C3 serum level. Thirty-nine (68.4%) patients had serum auto-antibodies (14 patients had auto-antibodies against one ribonucleoprotein (RNP), SSA52/TRIM21, SSA/Ro60, SSB or Sm and 25 patients against two or more of the above nuclear auto-antigens). Eighteen (31.6%) SLE patients had no detectable auto-antibodies. Serum IFN-α levels were elevated in 33 (58%) patients, with a median of 2 IU/mL (0-201).

Clinical involvement
Fever 16 (28) Weight loss or anorexia 12 (21) Lymphadenopathy 10 (17) Active cutaneous lupus 24 (42) Active lupus serositis 9 (16) Active lupus arthritis 21 (37) Active lupus nephropathy 11 (19) Proliferative nephropathy (classe III or IV) 4 (7) Membranous nephropathy (isolated classe V) 5 (9) Active neuropsychiatric lupus 4 (7) www.nature.com/scientificreports/ www.nature.com/scientificreports/ A3A and A3B expression was higher in patients with elevated serum IFN-α and was weakly correlated with IFN-α levels (for A3A: r = 0.3285, p = 0.0163 and for A3B: r = 0.3227, p = 0.0196), as expected (Fig. 1F,G). It is possible that the heterogeneity and fluctuation of symptoms associated with SLE partially contribute to the weak correlation between A3A and A3B levels and IFN-α. Although only A3A and A3B edit DNA, we also measured the expression of two other ISGs, A3G and RIG-I, as well as A3C. All were upregulated in SLE patients compared to healthy controls (A3G: p = 0.0354; A3C: p = 0.0270; and RIG-I: p = 0.0012) and particularly in patients with detectable serum IFN-I levels ( Fig. 1C-E). When SLE patients with detectable IFN-α in plasma were compared to patients with no detectable IFN-α, A3A (p < 0.05), A3C (p < 0.01), A3G (p < 0.01) and RIG-I (p < 0.01) expression were also more elevated (Fig. 1H-J). The observation that all were upregulated in SLE patients could be due to the fact that SLE is a multifactorial disease featuring many immunological dysregulations. Although A3C is generally not considered an ISG because it is much less responsive to IFN-I than A3G and A3F, its expression can be induced by IFN-I in PBMCs and in hepatocytes 15,16 . Furthermore, it is possible that other factors account for its upregulation in SLE patients. Nevertheless, since neither A3G nor A3C edit genomic DNA, it is unlikely that they contribute to lymphopenia. In line with this view, only A3A and A3B mRNA levels, were inversely correlated with lymphocyte counts (Fig. 1K,L) while A3C, A3G, and RIG-I mRNA levels were not (Fig. 1M-O), suggesting that genomic DNA editing by A3A and A3B may play a role in lymphopenia.
Taken together, these findings suggest that in SLE patients, the IFN-α response triggered by sustained cytosolic DNA due to oxidative mutations 39 leads to persistant A3A and A3B expression among other ISGs. These deaminases may participate in DNA catabolism, but most of all, they could edit nuclear DNA. In this case, uridine resulting from cytidine deamination is removed by Uracyl DNA glycosylases (UNG), creating an apyrimidinic site. This mobilization can activate the DNA repair machinery, leading to C > T or C > G mutations which may be fixed [5][6][7] . Alternatively, juxtaposed apyrimidinic sites will generate DSBs causing cell death 5,6 . In both cases, A3-mediated deamination events may generate neoantigens, increasing the B-cell response against self-epitopes in a feed-forward loop and contributing to lymphopenia. Accordingly, A3A mRNA levels were higher in SLE patients with auto-antibodies against nuclear proteins (p < 0.05) and against dsDNA ( Supplementary Fig. 1I) than in patients with no auto-antibodies, although statistical support was reached only for nuclear proteins. For A3B, mRNA levels were higher in SLE patients than in controls regardless of the presence of auto-antibodies (Supplementary Fig. 1J). For A3C and A3G, the levels of expression did not differ between SLE patients with or without auto-antibodies against nuclear antigens or against DNA (Supplementary Fig. 1 1 K and 1L). Although all A3 enzyme mRNA measured in SLE patients were upregulated, these observations are strongly suggestive of a role for A3A-and A3B-induced DSBs in cell death and the induction of auto-antibodies in the pathogenesis of SLE.
Hierarchical clustering of patients based on disease severity showed significantly exacerbated A3A, A3B and A3C expression in patients with a postive SLEDAI (Supplementary Fig. 1A to 1D), as well as with the presence or severity of flares (Supplementary Fig. 1E to 1H). Notably, hydroxychloroquine, corticosteroids and immunosuppressive treatment failed to reverse elevated A3A or A3B levels back to background levels (Supplementary Fig. 1M to 1N), while the difference in A3C and A3G mRNA levels did not differ significantly between treated SLE patients, untreated SLE patients and controls (Supplementary Fig. 1O and 1P).
In the long run, sustained exposure to A3A and A3B mutational fuel could also generate oncogenic driver events 23,30-32 , providing a direct molecular rationale for the higher prevalence of certain tumors among SLE patients. We therefore searched for the presence of subclonal oncogenic driver mutations conforming to the preferred target for A3A and A3B, i.e. TpCpW 7,21,66 , in 2 oncogenes, akt1 and the TERT promoter. We obtained sufficient sequences for all controls and 47 SLE patients. The threshold for distinguishing true subclonal mutations from error rate was set at 1 base call per 100 sequenced nucleotides. With these settings, we did not detect APOBEC3-mutations in akt1 (not shown) nor in the TERT promoter (Supplementary Table 1). Because SLE patients in this cohort do not have cancer, we expected to find only a handful of APOBEC3-mutations, if any. Indeed, APOBEC3-mutations are generally not detected in non-cancerous, TP53-and UNG-positive cells 67 . It is however not excluded that subclonal mutations may be present at frequencies below the set threshold or elsewhere in the genome.
High IFN-I-induced A3A expression leads to SLE-cell death. To further assess the role of detectable IFN-α and A3A in lymphopenia, freshly isolated PBMCs from SLE patients and healthy controls were put in culture with IFN-α for 24 h, mimicking flares. This setting recapitulates the fact that in SLE patients, but also in other chronic inflammatory conditions such as multiple sclerosis, flares often arise following an infectious episode, which likely triggers IFN-α. Initially, PBMCs from 5 SLE patients and 4 healthy controls were isolated and immediately exposed to IFN-α. By 4 h, IFN-α triggered a ~ tenfold increase in A3A mRNA, which was sustained throughout the time of the experiment in healthy controls (Fig. 3A). In SLE patients in contrast, IFN-α treatment increased A3A expression by more than 3 orders of magnitude (p < 0.05) (Fig. 3A) www.nature.com/scientificreports/ suggesting that SLE cells had died or were undergoing apoptosis (Fig. 3A). A3A and housekeeping gene mRNA could be quantified normally in cells from healthy controls treated with IFN-α, indicating that IFN-α per se did not induce massive cell death at these concentrations. In line with the high A3A mRNA levels in SLE patients, we recorded markedly higher levels of DSBs (γ-H2AX staining) in PBMCs from these SLE patients compared to healthy controls after 4 h (p = 0.0095) and 16 h (p = 0.019) in culture (Fig. 3B), indicating that higher A3A mRNA levels translate functionnally into increased DSBs and cell death. Next, PBMCs from 7 SLE patients and 3 healthy controls were isolated, left to rest overnight before IFN-α treatment. In controls, A3A upregulation started around 8 h and reached only ~ 100-fold after 24 h (Fig. 3C). In SLE patients, IFN-α increased A3A mRNA by up to ~ 200-fold as early as 4 h post-exposure and the increase persisted at 8 and 24 h. Thus, A3A induction was much faster and stronger in SLE patients compared to healthy control cells even when cells were left to rest overnight after isolation (p < 0.05 at 8 h) (Fig. 3C).
These results nicely recapitulate the profound lymphopenia observed during flares in SLE patients, and point to A3A as a chief actor in the massive cell death that characterizes flares. They suggest that not only are basal A3A expression and DSBs persistently higher in SLE patients than in healthy individuals, but also that their cells are "primed" such that A3A expression is readily boosted to extremely high levels while it takes healthy cells much longer to trigger A3A to 100-fold lower levels. No cell could survive such high levels of A3A. This immediate and massive upregulation of A3A suggests the A3A promoter is poised in SLE cells.
High prevalence of the TERT rs2853669 A > G polymorphism in SLE patients. Although we did not detect de novo somatic mutations attributable to A3 cytidine deaminase activity in the TERT promoter of our cohort, we found a surprisingly high prevalence of the germline polymorphism rs2853669 A > G among  (C) PBMCs from SLE or healthy controls were isolated from blood, left to rest overnight and only then treated with IFN-α. Relative A3A expression was quantified by qRT-PCR and normalized to the geometric mean of housekeeping genes RPL13A and GAPDH according to the Pfaffl method 65 for each sample. The fold-change in IFN-α-treated compared to non treated control 1 is reported. *p < 0.05; **p < 0.01; ***p < 0.001. In contrast, only 2 of 10 healthy controls (20%) were homozygous for rs2853669 A > G while all others harbored at least one copy of the rs2853669 A allele (5 were heterozygous and 3 were homozygous) (Fig. 4A). The rs2853669 A > G is a common germline polymorphism which disrupts a prexisting Ets/TCF binding site located 245 bp upstream of the TERT TSS. Accordingly, the rs2853669 A > G polymorphism decreases TERT transcription in vitro. The rs2853669 A > G polymorphism has been investigated in cancers with TERT promoter mutations, where it may reverse their effect and cell immortalization 54-59 , but has not been investigated in inflammatory conditions to our knowledge. This germline polymorphism is not an APOBEC-mutation and accordingly, we found no relationship between the prevalence of the polymorphism and detectable IFN-α in serum (Fig. 4B) nor with A3A or A3B mRNA levels (Fig. 4C,D) or disease severity (severe or mild/moderate flares, SLEDAI > 10) (Supplementary Fig. 2A and 2B). Nevertheless, peripheral blood mononuclear subsets from SLE patients, and particularly terminally differentiated memory T cells, display shortened telomeres as a result of sustained immune activation. It was suggested that TERT activity is insufficient to compensate for telomere erosion and accelerated replicative senescence of immune cells in these patients [46][47][48][49]68 . Because the rs2853669 polymorphism decreases TERT activity, individuals carrying this polymorphim may have lower TERT activity and accelerated lymphocyte senescence. Although we could not test this hypothesis in our cohort, it is possible that repeated PBMC activation episodes during flares combined to a decreased inability to rescue activated immune cells from senescence could contribute to the frailty of SLE patients' PBMCs and to lymphopenia. www.nature.com/scientificreports/

Conclusions
Our results document that multiple A3 enzymes, including IFN-induced A3A, A3B and A3G, and A3C, are strongly and persistently upregulated in SLE patients despite immunosuppressive treatment, and suggest a link between elevated baseline A3A levels, the A3AΔ3B polymorphism and disease severity. Since A3A and A3B can access the nucleus, these enzymes have every chance of editing nuclear DNA leading to DSBs. In this context, exposure to IFN-α leads to an immediate and massive upregulation of A3A expression, far beyond levels compatible with DNA repair, and ultimately to cell death. It is easily conceivable how high steady state A3A together with shortened telomeres due in part to the presence of the rs2853669 A > G polymorphism, might prime cells such that any further inflammatory signal leads to impelling A3A expression and massive cell death. Massive cell death in turn leads to exposure of nuclear antigens, fueling the inflammatory response and further incrementing the generation of auto-antibodies in a vicious cycle. These findings further underscore the need to include therapies targeting interferons and/or specifically A3 enzymes in the management of patients with lupus and probably other inteferonopathies 42 .

Methods
Patients and patient samples. Blood 62,63,70 , and the therapeutic regimen were recorded on the day blood was drawn. The class of lupus nephritis was recorded according to ISN/RPS-2003 71 . Routine testing to determine anti-dsDNA Ab titers Farr test (Trinity Biotech; cutoff value: 9.0 IU/mL), anti-ribonucleoprotein Abs (anti-RNP, anti-Sm, anti-SSA/Ro60, anti-Ro52/TRIM21, anti-SSB [Luminex FIDI, Theradiag]), and laboratory analyses (complement C3 levels (Optilite, Binding Site), complete blood counts, serum creatinine, proteinuria and hematuria) were run. The presence of a severe or mild/ moderate lupus flare was recorded according to the SELENA-SLEDAI flare instrument 62 . For some analyses, patients were stratified according to the presence and severity of flares (i.e. no flare versus mild/moderate flare versus severe flare) or the SELENA-SLEDAI score as follows: SLEDAI = 0: non-active Lupus; 0 < SLEDAI ≤ 5: mild condition; 5 < SLEDAI ≤ 10: medium activity; SLEDAI > 10: severe activity. Serum-IFN-α biological activity, expressed in IU/mL, was determined by assessing the protection afforded by each patient's serum to cultured MDBK cells challenged with vesicular stomatitis virus (VSV), as previously described 72 . Bioassay sensitivity (i.e. the lower limit of detection) was 2 IU/mL. Serum-IFNα activity in healthy individuals is undetectable, i.e. < 2 IU/ mL 72 . The study was approved by CPP Ile-de-France VI Ethics Committee. Samples were collected between November 2015 and October 2016. Ten sex-matched healthy controls whose blood was collected between June 15 and July 26, 2016 by the ICAReB platform at Institut Pasteur (Paris, France) were used for all experiments except for cell culture. All the participants gave written informed consent in the frame of the healthy volunteers CoSIm-mGEn cohort (Clinical trials NCT 03,925,272), which was approved by the CPP Ile-de-France I Ethics Committee (Jan 18, 2011). For cell culture experiments, blood from healthy volunteer blood donors from the Croix-Rouge Luxembourg was used. Written informed consent was provided by all patients and healthy donors. The research was carried out in compliance with the Helsinki Declaration.
Blood (2 × 7 mL) was collected on EDTA or PaxGene. Peripheral blood mononuclear cells (PBMCs) were isolated immediately by centrifugation, washed, split in two and dry pellets were stored at − 20 °C for DNA extraction and at − 80 °C for RNA extraction until use.
RNA isolation and RT-qPCR. Total RNA was extracted from frozen PBMC pellets using Trizol. One microgram of total RNA was reverse transcribed with the QuantiTect reverse transcription kit (Qiagen). qPCR for A3A, A3B, A3C and A3G, RPL13A and GAPDH was performed in duplicate using Takyon Rox dTT Blue 2X Master Mix (Eurogentec), primers and Universal Probe Library probes as described in 16 or the Applied Biosystems kit for GAPDH. For RIG-I, the following primers were used: F RIG-I : 5′-CTT TTT CTC AAG TTC CTG TTGGA-3′ and R RIG-I :5′-TCC CAA CTT TCA ATG GCT TC-3′, with UPL probe #79 (Roche). All genes of interest were normalized to the geometric mean of housekeeping genes RPL13A and GAPDH according to the Pfaffl method 65 and the amplification efficiencies for these primers 16 . A3A, A3B, A3C, A3G and RIG-I relative expression levels were calculated by comparison with Healthy Donor C1.
Next generation sequencing (NGS). DNA was extracted from frozen PBMC pellets using the Epicentre kit and 1 μg of genomic DNA was amplified using Platinum HiFi Taq (Invitrogen). The Telomerase Reverse Trancriptase (TERT) core promoter was amplified using outer primers hTERT-out-F: 5′-AGT GGA TTC GCG GGC ACA GA 73 and hTERT-out-R: 5′-GGC TTC CCA CGT GCG CAG CAGGA 74 and nested PCR primers : hTERTin-F : 5′-GCA CCC GTC CTG CCC CTT CACCT and hTERT-in-R : 5′-CAG CAG GAC GCA GCG CTG CCTGA, spanning mutations C228T and C250T. Akt1 intron 1 + exon 2, spanning mutation E17K, was amplified using primers Akt1-F: 5′-GCT GCC TGG CGA AGG TCT GACG and Akt1-R: 5′-CCT TGT AGC CAA TGA AGG TGCC. PCR settings for both genes were: 5 min denaturation at 94 °C followed by 40 amplification cycles (94 °C 1 min, 63 °C 1 min and 68 °C 1 min) and a 10 min final extension at 68 °C. PCR products for the 2 genes were obtained for all controls and for 47 SLE patients for the TERT promoter and for 46 patients for Akt1. PCR products were gel purified, adaptors were added according to standard procedures for Illumina NextSeq 500 sequencing. For data processing, quality control was performed using FASTQC. All sequences were cut-off at a minimum quality PHRED score of 20. A minimum read length of 125 bp was selected. Reads were mapped against TERT