Oligomeric S100A4 Is Associated With Monocyte Innate Immune Memory and Bypass of Tolerance to Subsequent Stimulation With Lipopolysaccharides

Objectives: Most DAMPs in inflammatory diseases are TLR2- and TLR4-ligands and according to the current concept, repeated stimuli would result in tolerance. Aims of the study were to verify this assumption, to investigate whether epigenetic effectors are involved and to explore the situation in rheumatoid arthritis (RA). Methods: A trained immunity (TI) and tolerance protocol was established using peripheral blood monocytes from healthy donors, β-glucan and lipopolysaccharide (LPS). The training or tolerance capacities of RA-relevant DAMPs were tested. Results: β-Glucan-, oS100A4-, HMBG1-, and HSP90-pretreated monocytes showed increased IL-6 responses to LPS re-stimulation. β-Glucan, oS100A and tenascin C induced training of monocytes to release more TNFα. In comparison to β-glucan, most DAMPs tested induced less TI, with exception of oS100A4. Monocytes exposed to oS100A4 showed increased IL-1β, IL-6, and TNFα in response to LPS, in spite that both stimulate TLR4. RNASEq upon β-glucan or oS100A4 revealed similar changes in chemokines/cytokines and epigenetic effectors; 17 epigenetic effectors correlated with chemokine/cytokine gene expression; PRDM8 was associated with more chemokine and cytokine transcripts. Knockdown of PRDM8 abolished TI induced by oS100A4. In RA, plasma S100A4 correlated with increased CSF2, and increased PRDM8 transcription in RA monocytes was associated with increased plasma CCL5 and IL-6, as well as therapy-resistance. Conclusion: Bypass of tolerance by DAMPs might be a phenomenon as important as TI, since it could explain how chronic inflammation can be maintained in spite of an environment with multiple TLR2/TLR4-ligands. In RA monocytes, a PRDM8-dependent TI mechanism could be responsible for sustained chemokine/cytokines levels.


INTRODUCTION
Monocytes and macrophages play a central role in the pathophysiology of inflammation. For instance, in rheumatoid arthritis (RA), activated monocytes massively infiltrate synovial tissues and produce tumor necrosis factor-α (TNFα) (1)(2)(3). Accordingly, therapies aimed at blocking this cytokine have emerged as a major tool in the treatment of RA and other inflammatory diseases (4,5). These observations indicate the essential role of TNFα in the pathogenesis of RA. Indeed, RA synovial fibroblasts have an intrinsic ability to destroy cartilage and bone due to epigenetic changes (6), and TNFα further stimulates their aggressive behavior (7).
Recently, it was shown that innate immune cells during initial challenge with pathogens become trained and successively undergo long-term adaptive changes in their functional programs, resulting in altered gene expression and cell physiology upon re-stimulation with the pathogen (19)(20)(21). This de facto novel kind of immunological memory has been termed trained immunity (TI) or innate immune memory. An in vitro TI protocol has been developed to explore the mechanisms governing the TI phenomenon (19,22,23). It has been reported that CD14+ monocytes or PBMCs pre-treated with heat-inactivated Candida albicans or its most potent component, β-glucan, increases production of IL-6 and TNFα upon re-stimulation with different non-related microbial stimuli. Similar training properties to β-glucan has been implicated for many other PAMPs (23). Of great importance is that dependently on PRR engaged and dosage regimen applied, tolerance (i.e., decreased cytokine production) instead of training could be induced). The current concept predicts possible training upon stimulation of different receptors and pathways, while sequential stimulation of receptors triggering common pathways result in tolerance (23,24). Training or tolerance occur depending on the stimuli, sequence of stimulation, doses and signaling pathways.
TI in relation to endogenous danger stimuli has been shown for oxidized low density lipoproteins (25) and lipoprotein A (26). Possibly, alarm signals associated with tissue damage and sterile inflammation can induce TI through epigenetic regulation of transcriptional programs. A concern regarding TI in a disease such as RA is that mostly all circulating DAMPs are recognized by TLR2 and/or TLR4. According to the concept above, it could be expected that subsequent stimulation with LPS or repeated exposure with DAMPs will result in tolerance. We report here that oS100A4 shows important training abilities and induces cellular changes allowing to bypass tolerance to subsequent TLR4 ligands, including LPS. Here, we propose that histone methylase PRDM8 plays a pivotal role in DAMPs-induced TI. Such a mechanism might be relevant in chronic inflammatory diseases. It can be hypothesized that TI and bypass of tolerance in RA might explain the association with e.g., increased risk for atherosclerosis (27). Thus, elucidation of this pathway would potentially reveal novel pharmaceutical targets.

Preparation of Monocytes From Human Buffy Coats and Peripheral Blood Mononuclear Cells
In Figures 1, 2, buffy coats from healthy donors (n = 8, Red Cross, Schlieren, Switzerland) were diluted 5-times with PBS; 35 ml diluted blood were underlayed with 15 ml Ficoll-Paque Plus (GE Healthcare) and centrifuged at 400 g for 35 min. The cell layer (PBMCs) were washed in PBS, resuspended in RPMI+ (RPMI1640 containing GlutaMAX, HEPES, PenStrep and pyruvate) and counted. Three ml PBMCs (150 to 200 × 10 6 cells) were layered on 10 ml 48.5% hyperosmotic Percoll solution (48.5 ml Percoll Plus [GE Healthcare] diluted in 41.5 ml sterile water and 10 ml 1.6M NaCl) and centrifuged at 580 g for 15 min. The interphase was collected, washed in PBS, and centrifuged down at 400 g for 10 min. RMPI+ were added to the cell pellet and this suspension is referred as "monocyte-enriched suspension." The cells were counted and checked for viability.
In the other experiments (Figures 3-6, 8), EDTA peripheral blood of healthy donors were used to isolate PBMCs by Ficoll-Paque. PBMCs were washed 2-times in cold MACS buffer, centrifuged, resuspended in RPMI+ containing 1% human serum, and counted. Monocytes were isolated by the addition of 20 µl anti-CD14 beads/10 7 cells in 80 µl MACS buffer (Milteny Biotec). After an incubation of 15 min at 4 • C, the cells were washed in cold MACS buffer, centrifuged and resuspended in 50 µl MACS buffer per 10 7 cells, followed by magnetic separation. CD14+ cells were resuspended in RPMI+ including 1% human serum, counted and used for stimulation, as well as training and tolerance experiments.

Monocyte Training and Tolerance Experiments
Monocyte-enriched suspensions of 2 × 10 6 cells/ml were prepared in RPMI+ (n = 8). The cells were distributed in 96wells plate (200'000 cells in 100 µl per well) and incubated at 37 • C for 1 h to let them to adhere. Non-adherent cells were removed (mostly CD3+, CD14-T-cells). The cells were washed with warm PBS to remove the rest of non-adherent cells. Next, the cells were incubated in 200 ml RPMI+ including 1% human pooled serum at 37 • C for 24 h. This procedure was repeated two other times and results in a population of >95% CD14+ and < 5% CD3+ cells, as checked by flow cytometry. Training was performed similarly to previous protocols (19,22,23). Briefly, as positive controls, monocytes were trained with 1 µg/ml βglucan in RPMI+ containing 1% pooled human serum. RPMI+ with human serum served as a negative control. The cells were incubated for 24h at 37 • C, before changing the medium. In a first set (Figure 1, Supplementary Figure 1), increasing doses of DAMPs or defined doses of PAMPs (e.g., LPS, PAM3CSK4, flagellin) allowed to differentiate between doses resulting in training or tolerance. In a second set (Figures 2A,B), monocytes were exposed for 24 h tor increasing doses of S100A4/oS100A4 (1 to 10 µg/ml). A part of the cells were pre-treated with LPS (1 ng/ml) followed by 3 days rest before re-stimulation with increasing doses of oS100A4 (Figures 2C,D). Finally, a part of the cells were pretreated with increasing doses of oS100A4 followed by 3 days rest before re-stimulation with LPS (10 ng/ml) (Figures 2E,F).

RNA Extraction and Quantitative
Real-Time PCR (qRT-PCR) RNA was isolated from monocytes using the RNA MicroPrep kit (Zymo Res.), according to the manufacturer's instruction; qRT-PCR was done to confirm upregulated genes and to verify the effect of siRNA. Reverse transcription was performed using 10 nM dNTP, RNase inhibitor, random hexamer, Multiscribe Reverse Transcriptase and GenAmp PCR Buffer II (all from Thermo Fisher). qRT-PCR was conducted on an Applied Biosystems 7900HT Fast Real Time PCR System, using FastStart Universal SyBr Green Master Mix (Roche) (primer pairs listed in Table 1).

Transcriptome Arrays
Five samples with 3 conditions were selected (healthy male blood donors 43 ± 11 years old). Conditions included unstimulated controls and stimulation with 1 µg/ml β-glucan or 2 µg/ml oS100A4. Adjusted volumes and cell densities for 12-well plate were 1.2 x 10 6 cells/well and 1.2 ml medium/well. After incubation at 37 • C for 24 h, cell culture supernatants were collected. Integrity of RNA was evaluated by agarose gel electrophoresis using an Agilent Bioanalyzer. RNA quality was controlled by RNA electropherograms and by using the RNA integrity number (RIN) resulting from an Agilent software algorithm. In cooperation with the Functional Genomic Center Zurich, poly A mRNA selection was performed using magnetic beads before cDNA synthesis. The cDNA libraries were sequenced using the Illumina Hiseq Sequencer. Generated were alignment QC, transcripts' expression quantification and differential expression analysis between conditions. Filtering of low read counts was applied, eliminating weakly or sporadically expressed genes. Selected were normalized genes up-regulated by >2-times (mean and median).
Transfection CD14+ monocytes (n = 4 healthy donors) were seeded into 24-wells plates. Mock or siRNA against PRDM8 transcripts (Thermo-Fisher AM16708/133027) transfections were performed using lipofectamine 2000, as described by the manufacturer (Thermo-Fisher). Two µg/ml oS100A4 were added after 24 h, corresponding to the time of maximal PRDM8 mRNA knockdown. Cell culture supernatants were collected after 3 days and the medium was replaced by RPMI+ containing FIGURE 2 | Effects of S100A4 and LPS. (A,B) Dose-response curves of homodimeric S100A4/oligomeric oS100A4 regarding increased releases of IL-6 and TNFα into the cell culture supernatant. (C,D) The effect of LPS pre-treatment (1 ng/ml) on the response to oS100A4 (1-10 µg/ml). (E,F) the effect of oS100A4-pre-treatement (1-10 µg/ml) on the response to LPS (10 ng/ml) (n = 8 healthy donors, Mann-Whitney U-test, *p < 0.01 vs. controls without first stimulus). LOED: lowest effective dose, EC70: refers to the concentration of a compound, which induces a response 70% between the baseline and maximum after 24 h.
1% human serum and 10 ng/ml LPS. After 24 h of incubation, cell culture supernatant was collected and analyzed by ELISA.

RA Plasma and Monocytes Cohort
EDTA-plasma and peripheral blood monocytes were collected from 36 RA outpatients (58 ± 10 years old) visiting the Clinic of Rheumatology at University Hospital, Zurich, as well as from 18 healthy controls (56 ± 12 years old), and RNA was prepared as previously described. The local ethical commission accepted the protocol and patients gave their informed consent (Basec 2019-0015). The samples and related clinical data were anonymized. All patients fulfilled the ACR/EULAR 2010 classification criteria for RA. In average, RA patients had an ESR of 5.0 ± 4.8 mm/h, CRP of 13.6 ± 15.1 mg/L and a DAS28 of 2.8 ± 1.2. The seropositivity for IgM-RF and anti-CCP positivity were 81 and 78% (78% were positive for both and 5% were seronegative). Treatments included Rituxumab (76% of patients), Methotrexate (41%), Tocilizumab (34%), and Leflunomid (24%). Remission was defined as a DAS28 score < 2.6, combined with BSR < 20 mm/h and CRP < 5 mg/l.

Statistics
Mann-Whitney U-test was used for comparison between groups and Wilcoxon signed-rank test between control and stimulations. Spearman's rank correlation analysis allowed finding possible relationships between variables. Heat-maps using Z-scores were constructed using http://www.heatmapper.ca/expression/ after normalization to β-actin transcription. Venn diagrams were used to illustrate and select epigenetic effectors associated with chemokines/cytokines gene expressions. Two component (C1, C2) analysis was performed using means of correlation coefficients for given chemokine/cytokines and epigenetic effectors. This allowed to clearly distinguish clusters.
Dose-Response Curves Upon S100A4 or oS100A4 In a second set of experiments using monocytes isolated from 8 additional buffy coats, we focused on the effects of oS100A4. As expected, oS100A4, compared to its homodimeric form, showed a much higher efficiency to induced pro-inflammatory cytokines, as shown by the release of IL-6 and TNFα after 24 h (Figures 2A,B).

Sequences With LPS and oS100A4
Monocytes were pre-treated with LPS (1 ng/ml for 24 h) and 3 days later re-stimulated with increasing doses of oS100A4 (Figures 2C,D). Under such conditions, monocytes showed an oS100A4 dose-dependent induction of tolerance. Next, we investigated the dose-dependent effect of pre-treatment with oS100A4 on the subsequent response to LPS (Figures 2E,F). Monocytes were exposed to different concentrations of oS100A4, the medium was changed after 24 h, and after 3 more days, LPS was added. In contrast to the previous sequence, oS100A4primed monocytes showed increased IL-6 and TNFα in response to LPS. Regarding IL-6, training occurred also at higher doses (4-10 µg/ml) without induction of tolerance ( Figure 2E). For TNFα, a tight dose window was associated with training (2-4 µg/ml), the effect being abolished at higher doses ( Figure 2F).

Confirmation Experiments
In a third set of experiments, we investigated different combinations between β-glucan (1 µg/ml as first or 10 µg/ml as second stimulus), oS100A4 (2 µg/ml as first or 5 µg/ml as second stimulus) and LPS (1 ng/ml as first or 10 ng/ml as second stimulus). β-Glucan trained monocytes showed increased IL-6 and TNFα responses to oS100A4, similarly as they would respond to LPS. As observed in the previous sets, LPS-pre-treated monocytes induce tolerance when challenged by oS100A4. Of note, oS100A4 increased the release of IL-1β and CSF2, but these cytokines were not affected by pre-treatments with β-glucan or LPS, i.e. monocytes became not trained or tolerant for those readouts. More important were the effects of oS100A4 as training stimulus. oS100A4-primed monocytes potentiated the effect of β-glucan, similarly to LPS-pre-treated cells. This included increases of IL-6, TNFα and IL-1β, but not CSF2 which was already maximally stimulated by oS100A4. It was confirmed that oS100A4-priming did not induce tolerance to LPS. At opposite, this sequence and combination even significantly increased the release of IL-1β and IL-6 (changes of TNFα levels in this set were not significant). oS100A4 re-challenged doesn't induce tolerance to itself (Supplementary Figure 2).

β-Glucan-Stimulated Chemokines/Cytokines
Quantitative RT-PCR revealed upregulated transcription of CCL20, CXCL1, CXCL8, CSF2, IL1B, and IL6 by 3.5 to 12-times (Supplementary Figures 3A,B). CCL5 and TNF mRNAs were increased by 2-times only. A proteome array provided information regarding released factors after 24 h stimulation of monocytes with training doses of β-glucan (Supplementary Figure 4B). Only an increased release of CCL5 was detected. This was confirmed by ELISA.
The proteome array showed that oS100A4 had a much broader effect than β-glucan, with increased releases of CCL5, CCL20, CXCL1, CXCL5, CSF2, IL-1β, and IL-6, while TNFα levels in the cell culture supernatant was only minimally changed (Supplementary Figure 4C). This was confirmed by ELISA.

RNA Sequencing
RNASeq provided more information about global gene expression changes upon β-glucan or oS100A4. The following analysis is based on genes that were upregulated 24 h after stimulation with 1 µg/ml β-glucan or 2 µg/ml oS100A.
Correlation heat-maps (Figures 4A-D) allowed to recognize a cluster of 17 epigenetic effectors mRNAs associated with the transcription of chemokines or cytokines ( Figure 5A). The Venn diagrams allow logical operations and revealed relationships of epigenetic effectors with chemokines and cytokines (Figure 5B), suggesting that KMT2E and PRDM8 might be major regulators of pro-inflammatory mediators in both conditions. For simplification, the ranges highlighted (white) in Figure 4 are called "KMT2E/PRDM8 clusters." Component analysis allowed using correlation coefficients to clearly distinguish clusters of chemokines/cytokines that could be associated with specific epigenetic effectors (Figures 5C-F). For instance, both IL6 and TNF belong to the KMT2E/PRDM8 cluster, but TNF might also be regulated by effectors included in C2. Upon oS100A4, some chemokines appeared in the left quadrants, suggesting that they are down-regulated by those epigenetic effectors or regulated by other factors.
Since KMT2E (MLL5) function might be more a regulator of cell cycle progression than an epigenetic effector, we focused on histone methylase PRDM8.

PRDM8 Knockdown and Training
Twenty-four hours of after transfection, siRNA directed against PRDM8 transcripts reduced PRDM8 mRNA by 74% and intracellular protein level by 55% (Figures 6A,B). Next, we tested the effect of PRDM8 knockdown in a protocol of TI (Figures 6C,D). Transfection was performed after isolation; this was followed next day by exposure to oS100A4 for 24 h; after additional 3 days, medium was replaced including LPS. PRMD8 knockdown significantly decreased the oS100A4-induced release of cytokines over the first 3 days. Thereafter, oS100A4-primed monocytes showed significantly increased responses to LPS, reflecting a TI phenomenon. PRDM8 knockdown increased the response of control monocytes to LPS. However, in the context of oS100A4-primed monocytes, PRDM8 knockdown abolished TI in response to LPS. These data clearly associate epigenetic modulator PRDM8 with oS100A4-driven TI pathways.

S100A4 and CSF2 in RA
Patients with high S100A4 levels showed more CSF2 and CRP. In RA patients, a significant relationship occurred between S100A4 and CSF2 ( Figure 7A).

PRDM8, CCL5, and IL-6 in RA
qRT-PCR showed no significant difference in transcription of PRDM8 between healthy donors and RA patients. Nevertheless, in RA, PRDM8 transcription positively correlated with plasma levels of CCL5 ( Figure 7B). In RA patients not receiving Tocilizumab, PRDM8 transcription positively correlated with plasma levels of IL-6 ( Figure 7C). Such relationships were not observed in healthy donors. In our patient cohort, Tocilizumab was associated with higher plasma levels of IL-6 and CCL20, but lower CRP.

Lower PRDM8 and Disease Remission
The 36 RA patients were divided into 2 groups, depending on their BSR, CRP, and DAS28 values. No remission was defined as fulfilling 2/3 following criteria: BSR > 20 mm/h, CRP > 5 mg/L and/or DAS28 score > 2.6. Taken all therapies together, poor responders showed significantly higher PRDM8 transcription (i.e., lower dCt values) than patients in remission (Figure 7D).

Differences Between β-glucan and oS100A4
Among differences between β-glucan and oS100A4, three important fact have to be noted (Figures 8A-D). Firstly, as expected, training with β-glucan increased LPS-induced transcription of IL6 and TNF. However, this is not the case in oS100A4-primed monocytes (Figure 8A). An alternative explanation for the increased release of pro-inflammatory cytokines in oS100A4-primed monocytes upon exposure to LPS regards modifications of the cytoskeleton, ion channels and secretory machinery. Thus, scinderin (Figures 8B,C) was downregulated, while numerous ion channels (example of a Rtype Ca ++ channel Figure 8D) were upregulated. The effect of SNX482 on THP-1 cells indeed suggested that such type of Ca ++ channel play a role in cytokine release upon stimulation (Supplementary Figure 6).

Most DAMPs Induced Weak Training
In this study, we showed for the first time that RA-related DAMPs, particularly oS100A4, can induce TI via epigenetic reprogramming involving histone methylase PRDM8. This mechanism might contribute to the well-known association between RA and e.g., cardiovascular diseases and provide a novel target for preventive pharmacotherapy.
Pre-treatment with DAMPs induced only weak training in comparison to β-glucan, with exception of oS100A4, for both IL-6 and TNFα. At higher doses, DAMPs can induce a moderate tolerance, but not comparable to endotoxin/PAMPs restimulations. The overall effect is dependent on the DAMP, doses and read-out.  Compared to controls (Scr), the response to LPS was significantly increased in oS100A4 (+Scr)-pretreated monocytes (arrow). sIRNA against PRDM8 increased the response to LPS in untreated cells. However, in the context of oS100A4-primed cells, the opposite occurs; TI was abolished and the response to LPS was strongly reduced. The discrepancy between up-regulation of transcripts and missing release could be due at least in part to increased degradation in proteasomes (28,29).

oS100A4 Priming and Trained Immunity
As a second stimulus, oS100A4 produced similar responses as LPS. Thus, in β-glucan-trained monocytes, it increases release of IL-6 and TNFα, while in LPS-pre-treated cells releases upon oS100A4 of those cytokines were drastically inhibited. Most importantly, oS100A4-pretreated monocytes not only showed a lack of tolerance to LPS, but they even increased releases of IL-6, TNFα and IL-1β in response to LPS. A S100A4-mediated increased response to LPS can be relevant in pathogenesis (30). Of interest was the increased transcription and release of CSF2 upon oS100A4 exposure. In mice, CSF2 positively influences the TI phenomenon (31). A possibility is that CSF2 enhances TNFα production, not before but after additional LPS stimulation (32).
Histone modifications are associated with transcriptional repression or activation, depending on the type of modification and locus modified. Methylation of lysine and arginine residues also can result in transcriptional activation. For instance, in rats, histone methylation is associated with persistent release of pro-inflammatory cytokines (35). Particularly interesting are the increased transcription of PR/SET domain proteins (PRDMs) encoding for histone methyltransferases. PRDMs are implicated in fundamental aspects of cellular differentiation (36,37). Upregulated PRDM8 transcription correlated with 31 chemokines/cytokines mRNAs, including CCL5, CCL20, CSF2, IL6, and TNFA. The importance of PRDM8 is emphasized by its central position in the Venn diagram, suggesting a key role in chemokine and cytokine regulation for both stimuli. Very few biological functions are known for PRDM8, with exception that in mice fibroblasts it methylates histone H3K9 and exerts a negative regulatory role on murine steroidogenesis (38); in addition, in mice, it is required for normal development of the neocortex (39). Thus, it was important to clarify its role during the process of TI.
Histone Demethylase PRDM8 as a Switch in Trained Immunity PRDM8 knockdown increased the release of IL-6 and TNFα after stimulation with LPS. Thus, in the context of an endotoxin stimulation, PRDM8 may limit the pro-inflammatory response, as expected from a histone methylase. However, the situation is completely different in oS100A4-primed monocytes; siRNA against PRDM8 not only decreased the release of IL-6 and TNFα upon oS100A4, they also abolished oS100A4-induced TI. In this sequence, the effect of oS100A4 dominates over LPS, probably due to global chromatin modifications. PRDMs either act as direct histone methyltransferases or recruit a suite of histonemodifying enzymes to target promoters (37). Previously, it has been shown in mice that a histone demethylase inhibitor (but not a histone methylase inhibitor) abolished TI (19). A limitation of the study could be the fact that monocytes from healthy donors, instead of RA treatment naïve patients were used. In patients, pre-existing chromatin changes or gene polymorphism could affect TI. Another limitation is that our analysis considers the response of monocytes to single exposure with β-glucan or oS100A4, reflecting how the cells are modified before they respond to a second stimulus. Interestingly, additional observations suggested that the mechanisms leading to an increased release of pro-inflammatory cytokines in fact are different between β-glucan-or oS100A4-pretreated monocytes. TI has been defined by increased release of pro-inflammatory cytokines into the culture supernatant and not necessarily by increased transcription. β-Glucan induces activation histone marks on the IL-6 and TNFα promoters, in particular H3K4 methylation (21). Thus, as expected, upon re-stimulation with LPS, β-glucan-trained monocytes increased the transcription of IL6 and TNF, i.e., higher than with LPS alone. This is followed by increased release. Monocytes primed with oS100A4 and re-stimulated with LPS, however, showed no more transcription of IL6 and TNF than cells exposed to LPS alone. An additional mechanism upon oS100A4 could be related to the downregulation of scinderin (SCIN), which controls the cortical membrane actin filament network. In addition, upregulation of ion channels might render monocytes more reactive to further stimuli; in particular CACNA1E, which encodes the alpha-1E subunit of the R-type Ca ++ channels. The fact that SNX482 (a potent and selective, voltagedependent R-type Ca ++ channel blocker) inhibits LPS-induced release of IL-6 and TNFα (Supplementary Figure 6) suggests that upregulation of Ca ++ channels could play a role in the increased response to a second stimulus. According to NCBI Genes, these channels mediate the entry of Ca ++ into excitable cells and are also involved in a variety of Ca ++dependent processes, including muscle contraction, hormone, or neurotransmitter release, gene expression, cell motility, cell division and cell death. It is important to note that such processes also can be regulated by epigenetics (40,41). Massive chromatin modifications occurred, as suggested by up-regulations of genes such as HMGA2, B2, or SMARCA1. Therefore, it will be interesting to investigate the effect of DAMPs on chromatin using a more global approach, such as ATAQ-Seq.

Monocyte Training in Rheumatoid Arthritis
In RA, elevated plasma S100A4 significantly correlated with increased CSF2, in accordance with the effect of oS100A4 on CSF2 release in vitro. In 31% of RA patients, the plasma level of S100A4 can reach 2 µg/ml (12); in 46% of the RA synovial fluid, it ranges from 2 to 7 µg/ml. The possibility that peripheral blood macrophages in RA are trained is an important issue in the context of chronic inflammation and comorbidity with cardiovascular diseases (2,22,27,42). In RA monocytes, gene expression of several proinflammatory cytokines are increased. They spontaneously showed higher releases of IL-1β and IL-6 (but not TNFα) (1). A recent study (43) reports unchanged ex vivo responses to LPS between healthy donors and RA patients; the authors didn't identify an increase in H3K4me3 at IL6 and TNF promoters and therefore rejected the TI hypothesis in RA. In fact, it is even surprising that RA monocytes are not tolerant to LPS, being confronted to an environment with multiple TLR2/TR4-ligands. Our data showed that DAMPs probably induced a form of TI through other mechanisms. More recently, a shift toward glycolysis has been reported in RA peripheral blood monocytes (2) that resembles trained monocytes (32,42). In RA, the level of PRDM8 mRNAs correlated with plasma levels of CCL5; this might be a sign that DAMPs influence the activity of monocytes. Another argument in favor of the TI hypothesis in RA monocytes is the positive association between PRDM8 and plasma levels of IL-6; this is not shown by healthy donors. Both the RNASeq correlation analysis upon stimulation and the PRDM8 knockdown experiment predicted a positive association between PRDM8 and IL-6 in trained monocytes. Finally, increased PRDM8 transcription was associated with therapy-resistance.
In RA, extracellular S100A4 released by activated synovial fibroblasts is involved in the paracrine regulation of several matrix-degrading enzymes and modulation of the transcriptional activation function of the tumor suppressor protein p53 (44). In S100A4-stimulated PMBCs, enhanced production of proinflammatory cytokines is at least partly mediated by activation of the transcription factor NF-κB and the MAP kinases p38 and ERK1/2 (13). S100A4 also influences the phenotype and activity of endothelial cells (45) and vascular smooth muscle cells (46). Thus, S100A4 could be involved in the comorbidity of RA such as cardiovascular diseases.
Finally, additional studies have to investigate whether autocrine/paracrine factors, such as S100A8/S100A9 or CSF2, play a role in DAMPs-induced TI.
In conclusion, even in the case that DAMPs act via TLR2/TLR4, this doesn't result in endotoxin tolerance. The bypass of tolerance might be a phenomenon as important as TI, since it could explain how chronic inflammation can be maintained. It also allows trained monocytes to respond adequately to infection. At concentrations measured in RA plasma, oS100A4 can induce TI. Interestingly, many similarities in chemokine/cytokine/epigenetic effectors and their relationships were found between β-glucan and oS100A4. The effect of β-glucan was explained through activating histone modifications at the IL6 and TNF loci and increased transcription upon second stimulus. This is less obvious regarding oS100A4-treated monocytes or RA monocytes. Nevertheless, the training effect of oS100A4 could be explained at different levels, including chromatin remodeling, together with changes in the cytoskeleton and secretory machinery (Figure 9). In association with epigenetic modifications, we suggested that upregulated histone methylase PRDM8 plays a key role in driving the cytokine/chemokine response. In RA monocytes, a PRDM8-dependent training mechanism could be responsible for sustained release of IL-6 and possibly other chemokines/cytokines. Through better understanding the effect of DAMPs on monocytes, novel therapeutic strategies could be developed, based on the inhibition of specific epigenetic mechanisms.

AUTHOR CONTRIBUTIONS
MN: conception and design of the work, analysis of data, wrote the manuscript. AP: acquisition and interpretation of data. NR, EL, and ES: design of the work, revising it critically for important intellectual content. LJ: design of the work, provided approval for publication of the content. MGN: conception and design of the work, revising it critically for important intellectual content. MG: conception and design of the work, acquisition and interpretation of data, revising it critically for important intellectual content. AC: acquisition and interpretation of data. OD: conception and design of the work, interpretation of data, revising it critically for important intellectual content. EK: conception and design of the work, acquisition and analysis of data, revising it critically for important intellectual content. KL: acquisition of data for the work, revising the new version critically, final approval. All authors provided approval for publication of the content, agree to be accountable for all aspects of the work.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.

2019.00791/full#supplementary-material
Supplementary Figure 1 | Examples of dose responses effects of DAMPs in the TI/tolerance protocol. Peripheral blood monocytes were exposed to increasing doses of DAMPs for 24 h and stimulated with 10 ng/ml LPS 3 days later (n = 8 healthy donors). IL-6 and TNFα were measured by ELISA in the cell culture supernatant. (A,B) 2.5-5.0 µg/ml oS100A4 showed TI for both IL-6 and TNFα. (C,D) 1.25-1.5 µg/ml HMGB1 induced TI more for IL-6 than TNFα. (E,F) 0.31-2.5 mg/ml tenascin C is associated with TI for TNFα only, at higher doses it produced tolerance. (G,H) HSP90 produced weak TI for both IL-6 and TNFα. The upper full line labeled βG represents the median TI obtained with 1 µg/ml β-glucan. The dotted line represents the baseline; the gray zone labeled LPS shows the zone of endotoxin tolerance, i.e., below the median obtained with repeated exposure to LPS (1 ng/ml followed by 10 ng/ml 3 days later).
Supplementary Figure 2 | Parts of the confirmation experiment regarding repeated exposures with the same stimulus. (A,B) β-Glucan by itself did not result in tolerance, in contrast to the repeated exposure to LPS. At opposite, after repeated stimulation with β-glucan, a detectable increase of TNFα release can occur. (C) Repeated oS100A4 stimulations didn't modify the release of cytokines and most importantly didn't result in tolerance. Peripheral blood monocytes from healthy donors were exposed to 1 µg/ml β-glucan, 2 µg/ml oS100A4 or 1 ng/ml LPS for 24 h and 3 days later were re-exposed to the same stimulus. Cytokines in the cell culture supernatant were measured by ELISA.