Mild COVID-19 imprints a long-term inflammatory eicosanoid- and chemokine memory in monocyte-derived macrophages

Monocyte-derived macrophages (MDM) drive the inflammatory response to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and they are a major source of eicosanoids in airway inflammation. Here we report that MDM from SARS-CoV-2-infected individuals with mild disease show an inflammatory transcriptional and metabolic imprint that lasts for at least 5 months after SARS-CoV-2 infection. MDM from convalescent SARS-CoV-2-infected individuals showed a downregulation of pro-resolving factors and an increased production of pro-inflammatory eicosanoids, particularly 5-lipoxygenase-derived leukotrienes. Leukotriene synthesis was further enhanced by glucocorticoids and remained elevated at 3–5 months, but had returned to baseline at 12 months post SARS-CoV-2 infection. Stimulation with SARS-CoV-2 spike protein or LPS triggered exaggerated prostanoid-, type I IFN-, and chemokine responses in post COVID-19 MDM. Thus, SARS-CoV-2 infection leaves an inflammatory imprint in the monocyte/ macrophage compartment that drives aberrant macrophage effector functions and eicosanoid metabolism, resulting in long-term immune aberrations in patients recovering from mild COVID-19.


INTRODUCTION
The Coronavirus disease 2019 (COVID-19) has emerged as a global pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infections 1 . Long-term symptoms of COVID-19 are common after severe disease 2 , but may also affect 15-20% of individuals with previous mild disease 3 . Monocyte-derived macrophages (MDM) drive the inflammatory response to SARS-CoV-2 and contribute to cytokine storms in severe COVID-19 4,5 . Severe COVID-19 is associated with profound changes in the myeloid compartment, including expansion of dysfunctional, proinflammatory monocytes during the first weeks after SARS-CoV-2 infection 6,7 .
Eicosanoids are bioactive metabolites of polyunsaturated fatty acids (PUFAs) with key roles in infection and inflammation 8 . Eicosanoids are formed from arachidonic acid (AA) through different enzymatic pathways, including the cyclooxygenase (COX) pathway, synthesizing prostanoids and the 5-lipoxygenase (5-LOX) pathway, generating leukotrienes (LTs) 8 . LTs are potent granulocyte-chemotactic metabolites which cause bronchoconstriction, vascular leakage, and airway remodeling 9 . Resident and recruited macrophages in the lung produce high levels of cysteinyl LTs (cysLTs) and leukotriene B 4 (LTB 4 ), thereby promoting granulocyte infiltration, airway inflammation and tissue remodeling 8 . Serum and airway prostanoid-and LT levels are increased in severe COVID-19 10,11 , suggesting a role for eicosanoids in the immune response to SARS-CoV-2 infection.
By studying transcriptome-and lipid mediator profiles in MDM of convalescent SARS-CoV-2-infected individuals with previous mild disease, we show that inflammatory gene expression and eicosanoid profiles as well as altered responsiveness to inflammatory cues are maintained at 3-5 months post infection as well as throughout macrophage differentiation. Pro-inflammatory 5-LOX metabolites were selectively increased in post COVID-19 MDM, suggesting that SARS-CoV-2 infection drives a pro-inflammatory eicosanoid reprogramming that contributes to long-term alterations in innate immune cell function.

RESULTS AND DISCUSSION
Recent studies have identified immunological changes in individuals recovering from severe or moderate acute COVID-19 for up to 12 weeks post infection 6,7,12,13 ; however potential immune aberrations in the majority of SARS-CoV-2-infected patients, affected by mild disease, have remained obscure.
Monocyte-derived macrophages of convalescent COVID-19 patients show pro-inflammatory transcriptional reprogramming and enhanced LPS responses Our recent work had shown that patients suffering from chronic airway inflammation exhibit transcriptional reprogramming of MDM 14 , a cell type implicated in COVID-19 pathogenesis 7 . To investigate whether SARS-CoV-2 infection induces persistent changes in MDM gene expression, we studied a sub-cohort from a large SARS-CoV-2 seroprevalence study in healthcare workers 15 (Table S1, Figs. 1a, S1a). To mimic the pulmonary cytokine milieu, in which infiltrating monocytes differentiate into macrophages, MDM were differentiated in the presence of GM-CSF and TGF-β1 16,17 , which resulted in a similar MDM population in seronegative and seropositive subjects (Fig. S1b).
CCL2, which is increased in monocytes during severe, acute disease 6 , was upregulated in post COVID-19 monocytes, suggesting a persistent inflammatory imprint despite mild disease in the investigated cohort (Fig. 1b).
SARS-CoV-2 S-protein-triggered IFN response is exaggerated in post COVID-19 MDM To define consequences of SARS-CoV-2-induced macrophage reprogramming for re-infection or vaccination, we investigated the response of post COVID-19 MDM to SARS-CoV-2 spike (S)protein. Entry of SARS-CoV-2 is mainly mediated via recognition of its transmembrane S-glycoprotein by angiotensin-converting enzyme 2 (ACE2) and processing by TMPRSS2 29 . However, ACE2 and TMPRSS2 expression in MDM was 100 times lower compared to airway epithelial cells, the major cellular targets of SARS-CoV-2, regardless of inflammatory stimulation or glucocorticoid treatment ( Fig. S1d-f). Yet, macrophages can respond to S-proteins of SARS-CoV-1 or SARS-CoV-2 via innate sensing mechanisms including C-type lectins 30,31 , which were upregulated in post COVID-19 MDM (Fig. S1g).
Thus, previous SARS-CoV-2 infection imprints a proinflammatory macrophage phenotype, that mounts exaggerated chemokine-and IFN responses, but likely exhibits impaired T-cell stimulatory and pro-resolving capacities. This was in line with previous studies identifying a dysfunctional, pro-inflammatory monocyte activation for up to 12 weeks after SARS-CoV-2 infection 7,13 and additionally suggested the long-term persistence of a pro-inflammatory macrophage state following mild disease. Changes in gene expression of post COVID-19 MDM were amplified by inflammatory stimuli, suggesting a "trained" state that lasted for at least 5 months post infection. Mechanistically, this may be driven by IFN-mediated reprogramming as post COVID-19 MDM exhibited an exaggerated upregulation of multiple ISGs, including perforin-2 (MPEG1), a driver of type I IFN signaling 26 .
This suggested that the prominent synthesis of inflammatory eicosanoids is not limited to acute and severe COVID-19 10,11 and that reprogramming of innate immune cells may result in persistently enhanced LT production even following mild disease.
Of note, we did not analyze spontaneous eicosanoid production, but used Ca 2+ ionophore to elicit maximal eicosanoid responses, which allowed us to quantify lipid mediators in limited numbers of patient cells. Thus, eicosanoid profiles identified in the current study reflect a setting of acute inflammatory challenge. MDM of convalescent subjects also revealed a marked lower inferred soluble epoxide hydrolase activity. The epoxides of arachidonic acid have been reported to promote the resolution of inflammation, including mitigation of cytokine storms 37 . Accordingly, inhibition of the sEH has been proposed as a potential therapeutic target for COVID-19 38 . Our findings suggest that subsequent to mild COVID-19, MDM may exhibit a compensatory sEH activity that is shifted towards a pro-resolution state. In contrast to acute infection, which resulted in increased ALOX5 expression in neutrophils and monocytes 10 , we did not find evidence of increased 5-LOX pathway gene expression in post COVID-19 MDM (Fig. 3g). Instead, genes involved in upstream events of fatty acid and lipid mediator biosynthesis (e.g., FASN, DGAT2, PLA2G4C) were upregulated in post COVID-19 MDM compared to MDM from seronegative subjects, suggesting an MDM phenotype in position for rapid activation of lipid metabolic pathways.
Analysis of MDM eicosanoid profiles from donors of the same cohort at 12 months post infection indicated that LT and prostanoid synthesis of post COVID-19 MDM had largely returned to baseline levels at this time point (Fig. 3h, i). This suggested that pro-inflammatory eicosanoid reprogramming in mild COVID-19 is transient, but that it may contribute to an enhanced inflammatory propensity during the first months post SARS-CoV-2 infection.
When stratified into 5-LOX low-or high producers, post COVID-19 subjects with high MDM LT production exhibited less acute symptoms but a faster decline in SARS-CoV-2 specific IgG titers (Fig. 3j), indicative of an efficient acute anti-viral response 39 . However, the lack of a defined clinical diagnosis of long COVID and poor reporting of long-term symptoms in the studied post COVID-19 cohort, prevented us from establishing a clear link between high MDM LT production and long-term symptoms of SARS-CoV-2 infection. Thus, future studies should investigate eicosanoid reprogramming in a cohort with clinically defined long COVID. Such studies would be imperative to define a potential pathological relevance of the inflammatory macrophage memory observed in the current study.
As patients in our study were enrolled following seroconversion, we were not able to compare monocyte and macrophage profiles at 3-5 months post infection to those during acute disease. However, we observed a considerable overlap between transcriptional profiles of post COVID-19 MDM and published transcriptomes of macrophages from SARS-CoV-2-infected individuals with mild acute disease 40 . Thus, several of the DEGs identified in our analysis (MPEG1, CD163, CXCL9, MERTK, and MRC1) were increased and correlated with higher expression of 5-LOX pathway genes in mild vs. severe acute disease 40 . It will be important to compare macrophage reprogramming between convalescent COVID-19 patients with different disease severities as well as following infection with other respiratory viruses (e.g., influenza). While previous studies have suggested an acute and transient increase in eicosanoids during respiratory syncytial virus (RSV) or influenza A virus (IAV) infection [41][42][43] , a comprehensive assessment of macrophage eicosanoid profiles in these diseases is currently lacking. PGE 2 production was increased following IAV infection, however we did not observe increased PGE 2 production in post COVID-19 MDM. Similarly, transcriptional profiles of post COVID-19 MDM showed minimal overlap with post influenza macrophage gene expression profiles 44,45 , suggesting that infection with different respiratory viruses results in distinct macrophage reprogramming. Increased macrophage LTB 4 production may however contribute to protective immunity during acute infection with multiple respiratory viruses 41,43 . It will be important to determine, whether the persistent increase of LTB 4 may contribute to a decreased susceptibility to respiratory viral infection during the first months following SARS-CoV-2 infection.
As airway inflammation, including in COVID-19, is commonly treated by glucocorticoids, we investigated potential effects of glucocorticoids on LT synthesis by post COVID-19 MDM. Fluticasone propionate, a commonly used inhaled glucocorticoid, further increased LT synthesis by post COVID-19 at baseline or after stimulation with house dust mite (HDM), used as a ubiquitous trigger of airway inflammation (Fig. S2a-d). This suggested that glucocorticoid treatment may further aggravate the pro-inflammatory eicosanoid reprogramming in post COVID-19 subjects. Given the therapeutic efficacy of glucocorticoids in airway inflammation, the finding that glucocorticoids enhanced LT synthesis may be surprising. However, it is in keeping with studies showing no reduction in LTs following glucocorticoid treatment in humans or enhanced LT production following in vitro treatment with glucocorticoids 46-48 .

S-protein-triggered prostanoid response is enhanced in post COVID-19 MDM
To assess potential differences in eicosanoid production capacities under inflammatory conditions, we compared Ca 2+ ionophoreelicited eicosanoid production in post COVID-19 and seronegative MDM stimulated for 24 h with S-protein or LPS. S-protein stimulation profoundly altered eicosanoid profiles (Fig. 4a, b), provoking a prominent induction of prostanoids from the thromboxane pathway (TXB 2 and 12-HHTrE), while 5-LOX metabolites were reduced (Fig. 4b).
In contrast to eicosanoid profiles, cytokine production at baseline or following stimulation was not significantly different between post COVID-19 and seronegative MDM (Fig. S2e-h), suggesting that cytokine aberrations may not persist for >12 weeks or during monocyte-macrophage differentiation. However, in contrast to LTs, cytokine and prostanoid production by MDM was efficiently suppressed by fluticasone propionate (Fig. S2a-d, i). This suggested that cytokines and prostanoids are efficiently targeted, while exaggerated LT responses of post COVID-19 MDM are further exacerbated by glucocorticoids. Indeed, thromboxane is a major eicosanoid produced by inflammatory macrophages and involved in vascular and airway remodeling, thus its inhibition by glucocorticoids may provide a therapeutic benefit. However, glucocorticoids may in turn further enhance the heightened production of pro-inflammatory LTs by post COVID-19 MDM, thus promoting LT-driven airway inflammation and remodeling. Based on the enhanced production of 5-LOX-derived lipid mediators both in acute 10,11 and post-acute COVID-19 (this study), approved LT pathway inhibitors should be considered as regimens to treat and/ or prevent airway inflammation and remodeling during the first 6 months following SARS-CoV-2 infection.
Future studies should further decipher upstream receptors and epigenetic pathways that drive the persistent pro-inflammatory macrophage and eicosanoid reprogramming during SARS-CoV-2 infection. In addition, a potential heterogeneity in GM-CSF and TGF-β1-differentiated MDM from seronegative and seropositive individuals should be addressed in single cell analyses. LTs have been reported to induce CCL2 in monocytes 50,51 , suggesting that enhanced LT synthesis may drive exaggerated pro-inflammatory chemokine responses in post COVID-19 MDM. In turn, increased CCL2 production by post COVID-19 MDM or monocytes may promote LTB 4 production 52 . Thus, our combined RNAseq and LC-MS/MS data suggest a crosstalk between CCL2 and LTs, which perpetuates the persistent pro-inflammatory activation of monocytes and macrophages following SARS-CoV-2 infection. Due to limitations in patient material, we could not perform a comprehensive comparison of MDM and monocytes, however our data suggest that differences in CCL2 and fatty acid synthesis are at least partially present in undifferentiated post COVID-19 monocytes, which differentiate into inflammatory monocyte-derived macrophages when entering the lung 5 . The persistent upregulation of pro-inflammatory eicosanoids in post COVID-19 macrophages may have multiple consequences for subsequent immune responses, e.g. during bacterial or viral infection or in patients suffering from chronic inflammatory diseases such as asthma, thus requiring future investigation.

Study design
Symptoms of seronegative (SARS-CoV-2 seronegative) and post COVID-19 (SARS-CoV-2 seropositive) individuals were determined through a questionnaire in Spring 2020 and 3-5 months later, in Summer 2020. Percentage of each symptom was calculated separately for seropositive and -negative individuals (table S1). Sample sizes for each experiment are specified in the corresponding figure legends; an overview is depicted in Fig. S1a. All blood donors participated in the study after informed written consent.

Histology
For histology ALI cells were fixed in 4% formaldehyde and embedded in paraffin. Sections were cut and hematoxylin & eosin (H&E) stained at the Klinikum rechts der Isar, Dermatology Department.

RNA isolation
Cells were lysed in RLT buffer (Qiagen, Hilden, Germany) supplemented with 1% β-mercaptoethanol. RNA was extracted using a spin-column kit according to the manufacturer's instructions (Zymo Research, Freiburg, Germany) and transcribed into DNA using the HighCapacity cDNA Reverse Transcription kit according to the manufacturer's instructions (Applied Biosystems) or submitted for total RNA sequencing.

RNA sequencing
Library preparation was performed using the TruSeq Stranded mRNA Library Prep Kit (Illumina, San Diego, CA, USA). Briefly, RNA was isolated from MDM cell lysates according to the j IgG titers in serum or number of symptoms in MDM from post CoV donors stratified into 5-LOX low (z-score < 1) and high producers (z-score > 1). Bar graphs are depicted as mean + SEM. Statistical significance was determined by Mann-Whitney test. *p < 0.05; **p < 0.01; ***p < 0.001. manufacturer's instructions (Zymo Research). Total RNA quality and quantity was assessed by Qubit 4 Fluorometer (Invitrogen) and RNA integrity number (RIN) was determined with the Agilent 2100 BioAnalyzer (RNA 6000 Nano Kit, Agilent).
For library preparation, 1 μg of RNA was poly(A) selected, fragmented, and reverse transcribed with the Elute, Prime, Fragment Mix (Illumina). A-tailing, adaptor ligation, and library enrichment were performed as described in the TruSeq Stranded mRNA Sample Prep Guide (Illumina). RNA libraries were assessed for quality and quantity with the Agilent 2100 BioAnalyzer and the Quant-iT PicoGreen dsDNA Assay Kit (Life Technologies, Thermo Fisher Scientific). RNA libraries were sequenced as 150 bp pairedend runs on an Illumina NovaSeq 6000 platform. Sequencing was performed at the Helmholtz Zentrum München (HMGU) by the Genomics Core Facility.
Cytokine analysis (ELISA) Cell culture supernatants were analyzed for IL-6, IL-1β and IL-8 secretion using the human ELISA Sets (BD Biosciences, Franklin Lakes, NJ, USA) according to the manufacturer's instructions.

Lipid mediator quantification
Briefly, cell supernatants from 200,000 cells, stored in equal volume of methanol, were extracted using solid phase extraction (Evolute Express ABN, Biotage, Uppsala, Sweden) and lipid mediators (see Table S1) were quantified by liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS) 55 . Given that cell culture media has significant background levels of many lipid mediators, compounds whose concentration was below the media level were excluded from data analysis.
Real-time quantitative PCR 10 ng cDNA was used as a template. The list of applied primers (4 µmol/L, Metabion international AG, Planegg, Germany) can be found in the Supplement. FastStart Universal SYBR Green Master Mix (Roche, Basel, Switzerland) was used and fluorescence was measured on a ViiA7 TM Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific). The expression levels were normalized to the house-keeping genes GAPDH (for MDM), ACTB, HPRT1 and TFRC (average for NHBEs and ALI cultured cells). Relative gene expression was calculated as 2 ΔCT (ΔC T = C T (Housekeeper) -C T(Gene) ). For genes where expression could not be quantified, CT values were set to 40.
Data analysis and statistics LC-MS/MS and RNAseq data were analyzed using previously published procedures 14,48,55,56 . Details can be found in the Supplement.