Дц Юмис Форекс

дц юмис форекс

Google Scholar

Toll-like receptor signaling in thymic epithelium controls monocyte-derived dendritic cell recruitment and Treg generation

Introduction

The establishment of tolerance is a fundamental attribute of a healthy immune system. Since T cell antigen receptors (TCRs) are generated by random somatic recombination, i.e. could be self or nonself-specific, T cells that express a self-reactive TCR must be removed from the conventional T cell repertoire. The critical part of this process occurs in the thymic medulla where the strength of TCR recognition of self-antigens is probed by various types of antigen presenting cells (APCs), mainly dendritic cells (DCs), B-cells, and highly specialized medullary thymic epithelial cells (mTECs)¹. mTECs mediate the promiscuous expression of thousands of otherwise strict tissue-restricted self-antigens (TRAs), a large number of which are under the control of the transcriptional regulator Aire². The presentation of TRAs by mTECs can result in either the deletion of self-reactive T cells³ or their conversion into Tregs^4,5

It has been recently demonstrated that the process of cooperative antigen transfer (CAT) from mTECs to DCs is essential for the establishment of thymic tolerance^{6,7,8,9,10,11}. The complexity of CAT is foremost due to the heterogeneity of DCs in the thymus. These CD11c⁺ cells are comprised of two major categories: B⁺ plasmacytoid DCs (pDC) and classical DCs (cDCs), the latter which can be subdivided into Xcr1⁺CD8α⁺Sirpα^– classical type 1 DCs (cDC1) and Xcr1^–CD8α^–Sirpα⁺ classical type 2 DCs (cDC2)^12,13. While cDC1 arise primarily in the thymus, cDC2 and pDCs originate extrathymically and then migrate to the thymic medullary region^14,15. mTEC-derived antigens are transferred both to thymic resident cDC1^6,10 and cDC2^16,17. Although it has been shown that the migration of cDC1 and cDC2 to the vicinity of mTECs is affected by a gradient of Xcl1¹⁸ and Ccr2/Ccr7 ligands, respectively^19,20, the potential involvement of other chemokines in the regulation of CAT still awaits resolution.

Toll-like receptors (TLRs) sense various immunologically relevant microbial ligands such as lipoproteins, carbohydrates, and nucleic acids. All TLRs, with the exception of TLR3, signal through the adaptor protein, MyD88, which via the activation of the NF-κB pathway induces the expression of pro-inflammatory cytokines, chemokines, and other inflammation-related molecules²¹. While the exact role of non-canonical NF-κB signaling in the development and function of mTECs has been previously demonstrated^22,23,24, the impact of TLR signaling via the canonical NF-κB pathway in the physiology of mTECs remains undetermined.

Here, we show that, among TLRs, mTECs abundantly express TLR9, and the stimulation of which leads to the influx of Xcr1^–Sirpα⁺ cDC2 into the thymic medulla. RNA sequencing of stimulated mTECs reveals that the mechanism underpinning this phenomenon is related to the upregulation of a set of chemokines, whose receptors are predominantly expressed by a CD14⁺ subset of thymic DCs, which have been identified as monocyte-derived DCs (CD14⁺moDC). Furthermore, mice with MyDdeficient TECs, which exhibit a deficiency in the recruitment of CD14⁺moDC, also suffer from a decreased thymic Treg output and functionality, which renders the peripheral T cell repertoire prone to colitis induction.

Results

mTECs express a set of TLRs and signaling adaptors

The function of TLR signaling in the physiology of mTECs has not yet been studied in detail^25,26,27. We first determined that both mTECs^low and mTECs^high subsets (Fig. 1a and Supplementary Fig. 1a) expressed TLR2, 3, 4, and 9 (Fig. 1b). Remarkably, TLR9, which recognizes bacterial, viral or altered DNA²¹ and ligands associated with cellular stress²⁸, is highly expressed by mTECs^high at levels comparable to thymic cDCs (Fig. 1a, b and Supplementary Fig. 1b). Transcripts of TLR adaptors MyD88 and Trif²¹ were also readily detectable (Fig. 1c). Although the levels of TLR4 and TLR9 were higher in mTECs^high, the major producers of Aire, our analysis of Aire^+/+ and Aire^–/– mice revealed that TLRs are expressed in an Aire-independent manner (Fig. 1d).

a Gating strategy used for the analysis of TEC populations and general thymic conventional DCs. MACS enriched CD45^– and EpCAM⁺CD11c^– pre-gated cells were further divided into cTECs (Ly51⁺), mTECs^low (MHCII^lowCD80^low), and mTECs^high (MHCII^highCD80^high). Thymic conventional DCs were gated as CD11c⁺MHCII⁺ from the CD45⁺ fraction. A more detailed gating strategy is found in Supplementary Fig. 1a, b. b Representative flow cytometry histograms of TLR expression on mTECs and DCs isolated from the thymus (n = 3 independent experiments). cMyD88 and Trif mRNA expression is determined by qRT-PCR from FACS sorted mTECs and DCs. The expression is calculated relative to Casc3 and normalized to the highest value within each experiment=1 (mean ± SEM, n = 3 samples). d Representative flow cytometry histograms of TLR expression on mTECs from Aire^+/+ and Aire^–/– mice, (n = 3 independent experiments). e Representative comparative flow cytometry plots of different TEC subpopulations in MyD88^fl/fl and MyD88^ΔTECs mice. f Quantification of TEC frequencies from plots in e (mean ± SEM, n = 6 mice). Statistical analysis was performed by unpaired, two-tailed Student’s t-test, p-values are shown. ns = not significant.

Full size image

To assess the significance of TLR/MyD88 signaling in TECs development, we crossed a thymic epithelial cell-specific Foxn1^Cre driver²⁹ with a MyD88^fl/fl transgenic mice³⁰ (hereafter called MyD88^ΔTEC). In comparison to the control, MyD88^ΔTECs mice showed no significant differences in the frequency of all tested TEC subpopulations (Fig. 1e, f), suggesting that canonical NF-κB signaling through TLRs/MyD88 does not affect mTEC^high maturation. Similarly, in all mTEC^high subsets, the expression of CD80, CD86, PD-L1, CD40, and ICOSL on was not altered (Supplementary Fig. 1c).

Together, this data demonstrates that TLRs are broadly expressed by mTECs and MyDdependent signaling has no apparent impact on TEC subpopulation frequency.

MyDdependent chemokine expression in mTECs^high

Given the high expression of selected TLRs in mTECs^high cells, we assessed the impact of the absence of TLR signaling in unperturbed conditions. RNA-sequencing of mTECs^high (sorted as shown in Supplementary Fig. 1a) from wild type (MyD88^fl/fl) and MyD88^ΔTECs mice revealed MyDdependent transcriptional variance (Fig. 2a) defined by differentially expressed transcripts (Fig. 2b and Supplementary Data 1 and 2). While of these transcripts were induced and 97 repressed by MyD88, they were not enriched for Aire-dependent or Aire-independent TRA genes³¹ (Supplementary Fig. 2a, left panel). Consistent with the role of TLR/MyD88 signaling in epithelial cells²¹, we found several differentially expressed genes (DEGs) which fell into one of two categories: (i) Il1f6 and Csf2 cytokines, (ii) Ccl25, Ccl4, and Ccl24 chemokines. These mediators act through receptors that are primarily expressed by myeloid cells and DCs³². Specifically, IL36R, the receptor for IL1F6, is expressed by DCs and T cells³³ while Csf2r, the receptor for Csf2, is expressed mostly by monocytes, macrophages, and granulocytes³⁴. The Ccr9, the receptor for Ccl25, is expressed by both thymocytes and pDCs driving their migration into the thymus^14,35. Both Ccr5 (receptor for Ccl4) and Ccr3 (receptor for Ccl24) are expressed predominantly on granulocytes and DCs modulating their migration into inflamed tissues^32,36. qRT-PCR analysis confirmed MyDregulated expression of selected genes in mTECs^high (Fig. 2c). Since the TLRs were postulated to sense both microbial and endogenous molecules²¹, we examined which of them could potentially act as a trigger. The analysis of mRNA expression of MyDdependent cytokines and chemokines (Fig. 2b, c) in the mTEC^high population isolated from either Germ-free (GF) or specific-pathogen-free (SPF) mice was comparable (Supplementary Fig. 2b), indicating that these signals are likely of endogenous origin.

a Principal component analysis of bulk RNA-sequencing data from mTECs^high (sorted as in Supplementary Fig. 1a) derived from MyD88^fl/fl and MyD88^ΔTECs mice. Data represents the analysis of n = 5 samples for each condition. b Volcano plot analysis of RNA-sequencing data described in a. Fold-change cutoff of log2 = ±1,0 and p-value: are marked by dashed lines (also in d, e). Differentially expressed genes are depicted in black, genes of interest are in red, and other detected genes in grey. c qRT-PCR analysis of relative mRNA expression normalized to Casc3 of genes selected from b (mean ± SEM, n = 3 samples). d Fold-change fold-change plot of RNA-sequencing data from CpG ODN or PBS in vitro stimulated mTECs^high (sorted as in Supplementary Fig. 1a) from MyD88^+/+ and MyD88^–/– mice (n = 4 samples for each condition). Color code as in b. e Volcano plot analysis of RNA-sequencing data from d, comparing CpG ODN versus PBS in vitro stimulated mTECs^high from MyD88^+/+ mice. Statistical analysis for b, d and e was performed by Wald test, p-value cutoff: f, g qRT-PCR analysis of Cxcl1 and Ccl5 mRNA expression (normalized to Cacs3) from in vitro (mean ± SEM, n = 4 samples) and intrathymically (mean ± SEM, n = 6 mice), respectively, CpG ODN or PBS stimulated mTECs^high from indicated animals. Statistical analysis for c, f, and g was performed by unpaired, two-tailed Student’s t-test, p ≤ = *, p ≤ = **, ns not significant.

Full size image

Next, we assessed the response of mTECs to TLR/MyD88 stimulation. Given the high expression of TLR9 (Fig. 1b), we stimulated mTECs^high from MyDdeficient (MyD88^–/–) and WT (MyD88^+/+) mice in vitro with CpG oligodeoxynucleotides (CpG ODN) or PBS. RNA-sequencing revealed significant changes in the transcriptional profile only in MyD88^+/+ cells. Notably, DEGs were associated with TLR9 stimulation (Fig. 2d, e and Supplementary Data 3 and 4), and of these, were upregulated while were downregulated. However, the pattern of expression of TRA genes remained largely unchanged after in vitro CpG ODN stimulation (Supplementary Fig. 2a, right panel). Importantly, among the most upregulated DEGs were two sets of chemokines: (i) Cxcl1, 2, 3, and 5, which signal via the Cxcr2 receptor, expressed predominantly on neutrophils³⁷ and (ii) Ccl3, 5 and 20 which signal via various chemokine receptors, including Ccr1, 3, 5, 6 which are expressed mostly on myeloid cells³². Cytokines (Tnfα, Il-6, Il12a, Il1f6 and Csf2) and other genes (Cd40) were also found to be upregulated (Fig. 2e). The upregulation of Cxcl1 and Ccl5 chemokines after in vitro (Fig. 2f) as well as in vivo intrathymic TLR9 stimulation (Fig. 2g) was confirmed by qRT-PCR analysis. As shown in Supplementary Fig. 2c, repeated intraperitoneal (i.p.) injection of CpG ODN was insufficient for the upregulation of chemokines in mTECs^high. It is of note that in vitro stimulation of TLR4 on mTECs^high by LPS also resulted in the upregulation of the previously noted chemokines, albeit at a lower level (Supplementary Fig. 2d).

In addition to TLRs, MyD88 also conveys signals generated by IL-1 family cytokines, such as IL-1β, IL or IL³⁸. Even though the receptors for these cytokines are expressed by mTECs^high (Supplementary Fig. 3a), only in vitro stimulation with IL-1β lead to the upregulation of cytokines and chemokines induced by TLR9 stimulation (Supplementary Fig. 3b).

Besides chemokines and cytokines, TLR/MyD88 signaling in mTECs^high (Fig. 2b) also regulated the expression of molecules associated with cornified epithelial pathway³⁹ (Supplementary Data 1–4). This specifically relates to genes that are associated with post-Aire mTECs^40,41, such as Krt10, Krt77 and Flg2 (Supplementary Fig. 3c). Moreover, previously published data has shown the enhanced expression of Il1f6, Cxcl3 and Cxcl5 in post-Aire mTECs⁴². Thus, we enumerated the total numbers of Involucrin⁺EpCAM⁺ cells in the medullary region of the CpG ODN intrathymically stimulated thymus. We did not observe any changes in the frequency of general mTECs subsets (Supplementary Fig. 3d) although the total numbers of Involucrin⁺ post-Aire mTECs were significantly increased (Supplementary Fig. 3e, f).

Together, these results show that TLR/MyD88 signaling in mTECs under physiological or stimulatory conditions regulates the differentiation of mTEC^high cells into Involucrin⁺ post-Aire stage. This stage is associated with the expression of a set of chemokines that signal via an overlapping set of chemokine receptors that are primarily expressed by DCs³².

TLR9/MyD88 signaling in mTECs targets Sirpα⁺ cDC2

Migration of different DC subsets into the thymus is orchestrated by distinct chemokines^14,18,19. Thus, we next assessed which of these subsets would be the target for TLR9/MyDinduced chemokines in TECs. We sorted three main subsets of CD11c⁺MHCII⁺ thymic DCs: B⁺ pDC, Sirpα^–Xcr1⁺ cDC1, and Sirpα⁺Xcr1^– cDC2 (Supplementary Fig. 4a), along with Gr-1⁺ granulocytes, CD4 and CD8 single positive thymocytes and performed qRT-PCR analysis of the chemokine receptors indicated above. Remarkably, apart from granulocytes, the chemokine receptors Cxcr2, Ccr1, 3, 5, and 6 were mostly expressed by DCs, specifically by cDC2 and pDC (Fig. 3a). This prompted us to quantify the relative frequencies of all thymic DC subsets in MyD88^ΔTECs in comparison to WT (MyD88^fl/fl) mice. In unstimulated conditions, TEC-intrinsic MyD88 signaling did not change the total frequency of CD11c⁺MHCII⁺ DCs (Fig. 3b, left plot). However, we observed alterations in the frequencies of DC subsets. While cDC1 were increased, the frequencies of pDC and cDC2 were diminished in the MyD88^ΔTECs thymus (Fig. 3b). In contrast, FACS analysis of TLR9-stimulated thymi revealed a significant increase in cDC2 accompanied by decreased cDC1 in the thymus of WT (MyD88^fl/fl) (Fig. 3c and Supplementary Fig. 4b, c) but not MyD88^ΔTECs animals (Fig. 3c). The frequencies of pDC remained comparable under these two conditions. This demonstrates that the recruitment of cDC2 to the thymus is attributable specifically to TLR9 signaling in TECs (Fig. 3c and Supplementary Fig. 4b). In agreement with medullary localization of cDC2 (Supplementary Fig. 4d), microscopically examined thymi from WT mice stimulated with CpG ODN showed an enrichment of CD11c⁺Sirpα⁺ cDC2 exclusively in the keratinrich medullary region (Figs. 3d, e).

a qRT-PCR analysis of the relative mRNA expression (normalized to Casc3) of indicated chemokine receptors on FACS sorted populations of thymic DCs; pDC plasmacytoid DCs, cDC1 classical type 1 DC, cDC2 classical type 2 DC, Gr-1^high = neutrophils, CD4 T = CD4⁺, and CD8 T = CD8⁺ thymic T cells. Sorting protocol of thymic DC subsets is provided in Supplementary Fig. 4a. T cells were sorted as TCRβ⁺ and either CD4 or CD8 single positive (n = 2 independent experiments). b Comparative flow cytometry analysis of total DCs (Gr-1^–CD11c⁺MHCII⁺) and different thymic DC subpopulations between MyD88^fl/fl and MyD88^ΔTECs mice enumerated according to gating strategy shown in Supplementary Fig. 4a (mean ± SEM, n = 17 for MyD88^fl/fl and n = 18 for MyD88^ΔTECs mice). c Flow cytometry analysis of different thymic DC populations (gated as in Supplementary Fig. 4a) isolated from CpG ODN or PBS intrathymically stimulated MyD88^fl/fl or MyD88^ΔTECs mice (mean ± SEM, pDC graph: n = 7 for ODN^–MyD88^ΔTECs and n = 8 for other three displayed items; cDC1 graph: n = 7 for MyD88^fl/fl and ODN^–MyD88^ΔTECs and n = 8 for ODN⁺ MyD88^ΔTECs mice; cDC2 graph: n = 6 for ODN^–MyD88^fl/fl, n = 7 for ODN⁺MyD88^fl/fl and ODN^–MyD88^ΔTECs and n = 8 for ODN⁺MyD88^ΔTECs mice). Statistical analysis in b, c was performed by unpaired, two-tailed Student’s t-test, p ≤ = *, p ≤ = **, p ≤ ***, p < = ****, ns not significant. d Microscopic examinations of thymic sections isolated from CpG ODN or PBS intrathymically stimulated WT mice. Cryosections were stained with keratin 14 (white), Sirpα (red), and CD11c (green). Scale bar represents 50 μm. The white dashed line demarks keratin rich medulla. e Quantification of CD11c⁺Sirpα⁺ cells in the medullar or cortical region of the cryosections shown in d (mean ± SEM, n = 12 counted square unites per medullary region; n = 10 and n = 9 counted square unites per PBS- and ODN-treated cortical region, respectively. Data are derived from three independent experiments). Statistical analysis was performed by unpaired, two-tailed Student’s t-test, p ≤ = **, ns not significant.

Full size image

Together, this data suggests that MyDdriven chemokines expressed by mTECs^high, target receptors on thymic Sirpα⁺ cDC2 and mediate their recruitment to the thymic medulla in steady state and TLR9 stimulatory conditions.

TLR9/MyD88 signaling in mTECs recruits CD14⁺moDCs

Chemokine-dependent migration of DCs to the proximity of mTECs, which underpins the mechanisms of CAT¹⁸, has been shown to be essential for the presentation of mTEC-derived antigens by DCs^6,10. One prediction from the TEC-dependent TLR/MyDinduced influx of Sirpα⁺cDC2 to the thymic medulla is that the frequency of CAT to this subset would be enhanced.

To verify this prediction, we crossed Foxn1^Cre mice with ROSA26^TdTOMATO leading to TEC-specific, cytoplasmic expression of TdTOMATO (TdTOM) protein in the thymus. In agreement with a previous study⁹ and as shown in Supplementary Fig. 5a, we found two major populations of TdTOM⁺ cells: (i) a TdTOM^high EpCAM⁺ population which was CD45^– and represented TECs expressing TdTOM endogenously (Supplementary Fig. 5b); and (ii) a CD45⁺ TdTOM⁺ population comprised of mainly CD11c⁺ DCs (Supplementary Fig. 5a) which acquired TdTOM via CAT (Fig. 4a). Interestingly, these DCs were enriched for the EpCAM⁺ marker (Fig. 4b) which was likely co-transferred with TdTOM⁹. Bone marrow (BM) chimeras of lethally irradiated Foxn1^CreROSA26^TdTOMATO mice reconstituted with WT BM cells showed that around 6% of donor-derived DCs acquired TdTOM (Supplementary Fig. 5c–e). This formally demonstrates that TdTOM is transferred from TECs to DCs.

a Experimental design. b Flow cytometry heat-map analysis showing the intensity of TdTOM fluorescence among MACS TCRβ-depleted cells from the thymus of the Foxn1^CreROSA26^TdTOMATO mouse. c Representative flow cytometry plots comparing the frequency of TdTOM⁺CD11c⁺ cells in the thymic MACS-enriched CD11c⁺ cells between the WT (Foxn1^Cre) and Foxn1^CreROSA26^TdTOM mouse. Cells were pre-gated as live, singlets, and Gr-1^–. d Quantification of TdTOM⁺CD11c⁺ cells from c (mean ± SEM, n = 6 mice). Statistical analysis was performed by unpaired, two-tailed Student’s t-test, p < = ****. e Representative flow cytometry histograms showing the frequency of TdTOM⁺ cells among pDC, cDC1, and cDC2 (gated as in Supplementary Fig. 4a). Gray histograms = Foxn1^Cre (control) mice, red histograms = Foxn1^CreROSA26^TdTOM mice. f Quantification of frequencies of TdTOM⁺ DCs among the indicated DC subsets (mean ± SEM, n = 6 mice). g Representative images from the Imagestream analysis showing intracellular localization of transferred TdTOM in MHCII⁺CD11c⁺ DCs from the thymus of Foxn1^CreROSA26^TdTOMATO (n = measured cells).

Full size image

It has been previously documented that distinct subtypes of thymic DCs vary in their capacity to acquire antigens from TECs^6,10,11,16. Whereas CAT of TdTOM from TECs to cDC1 and cDC2 is very potent in the Foxn1^CreROSA26^TdTOMATO system, it is limited in the case of pDC (Fig. 4e, f). This result was also corroborated with the use of BM chimeras which were described above (Supplementary Fig. 5f). Flow cytometry imaging showed that transferred TdTOM in MHCII⁺CD11c⁺ DCs is localized intracellularly (Fig. 4g).

To determine the heterogeneity of all thymic DC subsets that participate in CAT, we performed single-cell RNA-sequencing (ddSEQ)⁴³ of Gr-1^–CD11c⁺TdTOM⁺ cells isolated from thymi of Foxn1^CreROSA26^TdTOMATO mice. Two-dimensional tSNE projection clustering analysis revealed five different clusters of TdTOM⁺ DCs (Fig. 5a). Based on their expression profiles and previously described signature genes of cells from mononuclear phagocyte system (MPS)⁴⁴, we designated the clusters in accordance with MPS nomenclature¹³: two cDC1 clusters (Batf3): a cDC1a (Ccl5 and Ccr7) and cDC1b (Cd8a, Itgae, Xcr1 and Ppt1)⁴⁵; cDC2 cluster (Sirpα, Mgl2 and Cda)¹², moDC cluster (Sirpα, Cd14, Itgam, Cx3cr1 and Ccr2)⁴⁶; and one pDC cluster (Bst2, Ccr9, Siglech, and Ly6d)¹⁴ (Fig. 5b and Supplementary Data 5). This data allowed the clustering of DCs which participate in CAT according to their specific surface markers (Supplementary Fig. 6a). As shown in Fig. 5b, the previously defined thymic moDC subpopulation shared several markers with both cDCs (Itgax, Itgam, Sirpα, and Irf4) and classical tissue resident macrophages (Lyz2, Mertk, and Mafb). Due to the high mRNA expression of molecules associated with antigen processing and presentation by moDC subpopulation (Supplementary Fig. 6b), we tested their capacity to present mTEC-derived antigens and activate antigen specific T cells. Specifically, thymic CD14⁺moDCs isolated from the Aire-HCO mouse model expressing influenza hemagglutinin (HA) under the control of Aire regulatory sequences⁴⁷, were co-cultivated with HA-specific CD4⁺ T cell hybridoma cells (A5) carrying a GFP-NFAT reporter⁴. While the result demonstrated that thymic CD14⁺moDCs can efficiently present mTEC-derived antigens to T cells (Supplementary Fig. 6c), it seems that their previous detection was obstructed by using the previously established gating strategy (Supplementary Fig. 4a), by which they are indistinguishable from a conventional Sirpα⁺ cDC2 subset.

a Two-dimensional tSNE plot from ddSEQ single-cell RNA-sequencing from FACS sorted Gr-1^–CD11c⁺TdTOM⁺ DCs from the thymus of Foxn1^CreROSA26^TdTOMATO mice. The color code represents different cell clusters based on the mRNA expression profile of each cell. b Heat-map analysis of the expression of signature genes determining each subset defined in a. c Heat-map analysis of the expression of chemokine receptors by DC subsets defined in a. d Quantification of TdTOM⁺CD11c⁺ DC subsets (defined as in Supplementary Fig. 4a) in CpG ODN or PBS intrathymically stimulated Foxn1^CreROSA26^TdTOMATO mice (representative flow cytometry plots are shown in Supplementary Fig. 6d, f) (mean ± SEM, n = 9 mice). Statistical analysis was performed using unpaired, two-tailed Student’s t-test, p ≤ = **, p ≤ ***, p < ****. e Representative flow cytometry tSNE analysis of TdTOM⁺CD11c⁺cell population in PBS or CpG ODN intrathymically stimulated Foxn1^CreROSA26^TdTOMATO mice. tSNE analysis was performed using FlowJO software, based on the FSC-A, SSC-A, CD11c, MHCII, Sirpα, Xcr1, B, Mgl2 and CD14 markers (n = 2 independent experiments). f Quantification of frequencies of TdTOM⁺CD14⁺moDC or TdTOM⁺Mgl2⁺cDC2 from CpG ODN or PBS intrathymically stimulated Foxn1^CreROSA26^TdTOMATO mice (representative flow cytometry plots are shown in Supplementary Fig 6g) (mean ± SEM, n = 4 mice). g Flow cytometry analysis comparing the frequency of cDC2 (Sirpα⁺Mgl2⁺) and moDC (Sirpα⁺CD14⁺) between MyD88^fl/fl and MyD88^ΔTECs mice (mean ± SEM, n = 6 mice). Total Sirpα⁺ DC population was gated as shown in Supplementary Fig 4a. Statistical analysis in f and g was performed by unpaired, two-tailed Student’s t-test, p ≤ = **, p < ****, ns not significant.

Full size image

Next, we determined which of the five defined thymic DC clusters expressed the receptors for TLR9/MyDinduced chemokines/cytokines from mTECs (Figs. 2b, d, e). The heat map analysis of chemokine receptors identified by ddSEQ analysis revealed that most of these receptors were expressed by the Sirpα⁺CD14⁺moDC cluster (Fig. 5c, left panel). Interestingly, each of the TdTOM⁺ DC clusters expressed a specific set of chemokine receptors (Fig. 5c).

Having characterized the CAT system with participating subsets of DCs in Foxn1^CreROSA26^TdTOMATO mice, we used it as a read-out to determine the targeting specificity of TEC-dependent TLR9/MyD88 stimulation on these DC subsets. First, in general, TLR9 intrathymic stimulation of Foxn1^CreROSA26^TdTOMATO mice boosted the frequency of total TdTOM⁺ CD11c⁺ DCs (Fig. 5d left graph and Supplementary Fig. 6d) as well as the mean fluorescent intensity (MFI) of TdTOM in these cells, demonstrating their enhanced rate of CAT under stimulatory conditions (Supplementary Fig. 6e). Second, as predicted, the observed increase in CAT was fully attributable to TdTOM⁺Sirpα⁺ DCs and not to other DCs populations (Fig. 5d right graph and Supplementary Fig. 6f). Third, and most importantly, the unsupervised flow cytometry tSNE analysis of the main DC subsets defined by markers revealed by ddSEQ analysis showed that the increase of TdTOM⁺ DCs was mostly due to the specific enrichment of CD14⁺moDCs (Figs. 5e, f), which also co-express chemokine receptors for ligands induced by TLR9/MyD88 signaling in mTECs (Figs. 2b, e and 5c). Concomitantly, we observed a decrease in Mgl2⁺ cDC2, Xcr1⁺ cDC1b, and B⁺ pDC (Fig. 5e, f and Supplementary Fig. 6g). Importantly, and further confirming the need of MyD88 signaling for its recruitment, the decreased frequency of total Sirpα⁺ DCs in the thymus of non-manipulated MyD88^ΔTECs mice (Fig. 3b) was shown to be accounted specifically by the diminishment of the CD14⁺moDC subset (Fig. 5g).

To find which of the chemokines described (in Fig. 2b, e) were responsible for CD14⁺moDC migration to the thymus, we crossed Cxcr2^fl/fl mice with the pan-hematopoietic driver Vav1^Cre to abrogate the signaling of its cognate ligands Cxcl1, 2, 3, and 5 that were among the most upregulated genes in mTECs after TLR9 stimulation. We observed no changes in the recruitment of CD14⁺moDC after TLR9 stimulation between Cxcr2^fl/flVav1^Cre and WT mice (Supplementary Fig. 6h). This suggests, that together with ligands of Ccr2, (i.e. Ccl2, 7, 8, and 12)¹⁹, the ligands of Ccr1, Ccr3 or Ccr5, or their combinations³⁶, regulate the entry of CD14⁺moDC into the thymic medulla.

Together, TLR9/MyDdependent chemokine signaling in mTECs specifically targets the recruitment and subsequent CAT from the mTECs to Sirpα⁺CD14⁺moDC subpopulation which exhibits a tangible capacity for antigen presentation.

TLR9/MyD88 signaling in mTECs affects Treg development

Previous studies have suggested that the development of thymic Tregs is dependent on antigen presentation by both mTECs and DCs^6,17,47. Specifically, antigen presentation by Sirpα⁺DCs¹⁷ and/or alternatively by CD8α⁺cDC1^6,10 was implied in the development of organ-specific Tregs. It has been also suggested that the increased ratio of Sirpα⁺DCs to CD8α⁺cDC1 leads to an enhanced production of thymic CD25⁺Foxp3⁺ Tregs^17,20. Since a decreased frequency of Sirpα⁺DCs (Fig. 3b), specifically CD14⁺moDCs (Fig. 5g) was observed in the thymus of MyD88^ΔTECs mice, we tested whether these effects would impact the development of the major thymocyte populations and Tregs. While the DN (CD8^–CD4^–), DP (CD8⁺CD4⁺), and CD8⁺ T cells frequencies were comparable between MyD88^ΔTEC and WT mice, CD4⁺ T cells, and more specifically CD25⁺Foxp3⁺ Tregs were significantly reduced in 4-week-old MyD88^ΔTECs mice (Fig. 6a–c and Supplementary Fig. 7a). Since it has been reported that in 4 week-old-mice nearly one half of CD25⁺Foxp3⁺ thymic cells consist of mature recirculating Tregs^48,49, we used CD73 protein staining to determine if Tregs reduced in MyD88^ΔTECs mice were newly generated (CD73^–) or recirculating (CD73⁺)⁵⁰. As shown in Fig. 6d, e, the abrogation of MyD88 signaling in mTECs affected mainly the generation of CD25⁺Foxp3⁺ thymic Tregs and not their recirculation. On the other hand, the CD25⁺Foxp3⁺ thymic Tregs were not reduced in newborn MyD88^ΔTECs (Supplementary Fig. 7b) or GF mice (Supplementary Fig. 7c) when compared to their WT SPF littermates. This, in association with unchanged chemokine expression in mTECs from GF mice, (Supplementary Fig. 2b) further strengthens the notion that the ligands that regulate the mTEC-mediated MyDdependent cellularity of Tregs is not likely of exogenous origin.

a Representative flow cytometry plots (left plot) and their quantification (right plot) comparing the frequencies of main thymic T cell populations between MyD88^fl/fl and MyD88^ΔTECs mice (mean ± SEM, n = 14 mice). b Representative flow cytometry plots comparing the frequencies of CD4⁺CD25⁺Foxp3⁺ thymic Tregs between MyD88^fl/fl and MyD88^ΔTECs mice. c Quantification of frequencies from b (mean ± SEM, n = 14 mice). d Representative flow cytometry histograms showing the expression of CD73 by CD4⁺CD25⁺Foxp3⁺ thymic Tregs (gated as in b). e Quantification of the total numbers of CD73^– and CD73⁺ thymic Tregs from d (mean ± SEM, n = 7 mice). f Quantification of the frequencies of thymic Tregs from CpG ODN or PBS intrathymically stimulated (7 days) MyD88^fl/fl or MyD88^ΔTECs mice (mean ± SEM, n = 4 for MyD88^ΔTECs and n = 9 for MyD88^fl/fl mice). g Quantification of the total numbers of CD73^– and CD73⁺ thymic Tregs from CpG ODN or PBS intrathymically stimulated (7 days) WT (C57Bl/6J) mice (mean ± SEM, n = 6 for ODN⁺ and n = 7 for ODN^– mice). h Quantification of frequencies of thymic Tregs from CpG ODN or PBS intrathymically stimulated (7 days) H2-Ab1^fl/fl or H2-Ab1^fl/flItgax^Cre (H2-Ab1^ΔDCs) mice (mean ± SEM, n = 6 for H2-Ab1^fl/fl and ODN⁺ H2-Ab1^ΔDCs and n = 7 for ODN^– H2-Ab1^ΔDCs mice). i Quantification of the total numbers of CD73^– and CD73⁺ thymic Tregs from CpG ODN or PBS intrathymically stimulated (7 days) H2-Ab1^ΔDCs mice (mean ± SEM, n = 6 for ODN⁺ and n = 7 for ODN^– mice). Statistical analysis in a, c, e–i was performed by unpaired, two-tailed Student’s t-test, p ≤ = *, p ≤ = **, p ≤ *** p < ****, ns not significant.

Full size image

To further explore the MyDdependent regulation of Tregs generation, we tested our prediction that TLR9/MyD88 stimulation of mTECs would lead to the opposite effect, i.e. boosted number of Tregs. Indeed, seven days after intrathymic injection of CpG ODN, we observed a significant increase in the frequency and total number of CD25⁺Foxp3⁺ thymic Tregs (Fig. 6f and Supplementary Fig. 7d–f). Importantly, this increase was completely dependent on TEC-intrinsic MyD88 signaling (Fig. 6f). Compared to the decreased numbers in CD73^– Tregs in MyD88^ΔTEC, intrathymic injection of CpG ODN led to increased numbers of not only CD73^– newly generated Tregs but also recirculating CD73⁺ Tregs (Fig. 6g and Supplementary Fig. 7g). This suggests that there are other mTEC-dependent mechanisms which after CpG ODN stimulation can affect the recirculation of Tregs into the thymus. One outstanding question related to the results from the above experiments (Figs. 3c and 5d–f and Supplementary Fig. 6c) is whether the increased generation of thymic CD25⁺Foxp3⁺CD73^– thymic Tregs is dependent on the antigen presenting capacity of DCs. To resolve this query, we intrathymically injected CpG ODN into H2-Ab1^fl/flItgax^Cre (H2-Ab1^ΔDCs) mice, where antigen presentation by DCs has been abrogated. As demonstrated in Fig. 6h, i and Supplementary Fig. 7h, the presentation of antigen by DCs is essential for the increase in numbers of newly generated CD73^–CD25⁺Foxp3⁺ thymic Tregs after TLR9 stimulation.

Next, we tested the physiological consequences of the decrease in production of Tregs in MyD88^ΔTECs mice. We took advantage of a T cell induced colitis model, where the adoptive transfer of naïve, Treg depleted CD4⁺ T cells into Rag1-deficient mice induces severe colitis⁵¹. In this experimental setup, and as illustrated in Fig. 7a, the i.p. injection of the CD4⁺ T cell population isolated from peripheral lymph nodes of either MyD88^ΔTEC or MyD88^fl/fl mice was compared to colitis-inducing transfer of CD4⁺CD45RB^highCD25^– cells isolated from WT mice.

a Experimental design of induced colitis. b Relative quantification of mice weight normalized to its value on day 0 (% of original weight) after T cell transfer over the time-course of the colitis experiment (mean ± SEM, n = 5–6 mice) Statistical analysis was performed by unpaired, two-tailed Student’s t-test comparing the relative weight of WT CD4⁺ with MyD88^ΔTECs CD4⁺ transferred mice (blue) or with WT CD4⁺CD45RB^highCD25^- transferred mice (red), p ≤ = *, ns not significant. c Representative H&E-stained slides of colon sections performed 8 weeks after T cell transfer. Scale bar represents μm (n = 5 for CD4⁺ WT and n = 6 for CD4⁺MyD88 ^ΔTECs and CD4⁺ No Tregs WT mice). d Relative quantification (normalized to average of control mice from each experiment) of colon weight/length ratio of T cell induced colitis experimental mice (mean ± SEM, n = 5–6 mice). e Relative quantification of the frequencies (normalized to average of control mice from each experiment) of CD4⁺CD25⁺Foxp3⁺ Tregs isolated from the spleens of experimental mice 8 weeks after T cell transfer (mean ± SEM, n = 5–6 mice). f Quantification of the Means fluorescent intensity (MFI) of CD25 protein expression in CD25⁺Foxp3⁺ Tregs (gated as in Fig. 6b) in MyD88^fl/fl and MyD88^ΔTECs mice (mean ± SEM, n = 12 mice) Statistical analysis in b, d–f was performed by unpaired, two-tailed Student’s t-test, p ≤ = *, p < ****, ns not significant. g Representative flow cytometry plots showing the frequency of proliferating OT-II T cells, co-cultivated with OVA pulsed BMDC and CD4⁺CD25⁺ Tregs cells (alternatively with CD4⁺CD25^– Tconv cells, black) isolated from LNs of MyD88^fl/fl (WT control, red) or MyD88^ΔTECs (blue) for 72 h. h Quantification of frequencies of proliferating OT-II Tcells form h (mean ± SEM, n = 4 wells from two independent experiments).

Full size image

Strikingly, mice that received CD4⁺ T cells from MyD88^ΔTECs began to lose weight ~4 weeks after adoptive transfer, behaving identically to the positive control. In contrast, mice that received CD4⁺ T cells from WT mice continuously gained weight over time (Fig. 7b). The clinical signs of colitis in mice receiving CD4⁺ T cells from MyD88^ΔTEC and in the positive controls were manifested by the presence of inflammatory infiltrates in the colon lamina propria, increased bowel wall thickness, presence of abscesses in colon tissue (Fig. 7c), increased spleen weight (Supplementary Fig. 8a, b), and a higher colon weight/length ratio (Fig. 7d and Supplementary Fig. 8a). To confirm the persistence of the transferred T cell population, we also analyzed Tregs frequencies in all conditions. We found that both positive controls and mice that received CD4⁺ T cells from MyD88^ΔTECs had severely diminished Tregs compared to WT controls (Fig. 7e). The very similar phenotype of mice that received CD4⁺ T cells from MyD88^ΔTECs and those which received CD4⁺CD45RB^highCD25^– suggested, that Tregs in MyD88^ΔTECs were not only reduced in numbers but also functionally altered. Along with the decreased expression of CD25 (Fig. 7f), Tregs from MyD88^ΔTECs mice showed a significantly reduced capacity to suppress the proliferation of OVA-specific OT-II T cells in vitro (Fig. 7g, h) and prevent the early onset of diabetes caused by activated KLGR1⁺ OT-I T cells in a RIP-OVA dependent autoimmune mouse model⁵² (Supplementary Fig. 8c–e).

Taken together, these results demonstrate that TLR/MyD88 signaling in TECs affects the development of thymic CD25⁺Foxp3⁺ Tregs. Specifically, in mice with MyDdeficient TECs, the frequency and functionality of thymic CD25⁺Foxp3⁺ Tregs was decreased and unable to prevent T cell induced colitis.

Discussion

Present study lends a support for the role of TLR signaling in the mechanism of central tolerance. First, we found that mTECs^high express TLRs, including TLR9, whose signaling is functionally wired to the expression of chemokines and genes associated with their post-Aire development. Second, the receptors for these chemokines are predominantly expressed by the Sirpα⁺ thymic population of CD14⁺moDCs whose enrichment in the thymus and subsequent CAT is positively regulated by mTEC-intrinsic TLR/MyD88 signaling. Third, TLR/MyD88 signaling in mTECs is important for the proper development of thymic CD73^–CD25⁺Foxp3⁺ Tregs since its abrogation resulted in a decreased number and the functionality of Tregs, associated with pathological effects in the mouse model of colitis.

The importance of TLR/MyD88 signaling in Aire-dependent autoimmunity was suggested in experiments conducted with MyD88^–/–Aire^–/– double-knockout mice. These mice develop more severe symptoms of autoimmunity than Aire^–/– single KO animals indicating the positive regulatory role of MyD88 signals in tolerance induction. Strikingly, neither the enhancement of MyD88 signals by an i.p. injection of TLR ligands, nor their diminishment in mice from GF conditions altered the severity of Aire-dependent autoimmunity⁵³. Our data advocates for a scenario in which the worsening of autoimmunity in MyD88^–/–Aire^–/– mice could be caused by the lack of MyD88 signaling in mTECs^high, downregulation of their chemokines needed to recruit CD14⁺moDCs and, consequently, suboptimal production of thymic Tregs. Consistent with the previous report⁵³, we confirmed that the extrathymically enhanced (i.p. CpG ODN) or the lack of bacterially-derived MyD88 signals (GF mice) had no effect on the expression level of these chemokines and cytokines in WT mice. This was further corroborated by the fact that GF mice displayed normal numbers of Tregs⁵⁰ (Supplementary Fig. 7c). This data demonstrates that the ligand triggering TLR9/MyD88 signaling in mTECs^high is likely of endogenous thymic-derived origin.

Since MyD88 also conveys signals from the receptors of IL-1 family cytokines (IL-1β, IL, IL)³⁸, we tested in vitro whether their signaling in mTECs^high could trigger chemokine responses similar to those observed upon TLR9 stimulation. Of this trio of cytokines, only IL-1β exhibited this capacity. This indicates that IL-1β could act as a co-regulator of chemokines and cytokine expression in mTECs^high. However, two observations suggest that TLR9/MyD88 signaling axis can act independently of IL-1β: (i) a direct, in vitro, stimulatory capacity of CpG ODN induces chemokine expression in sorted mTECs^high; and (ii) both in vivo intrathymic stimulation of TLR9/MyD88 signaling axis as well as its downregulation in MyD88^ΔTECs cells impacts the recruitment of the very same subsets of CD14⁺moDCs.

It has been postulated that Aire⁺ mTECs further differentiate into post-Aire cells, which downregulate the expression of MHCII and Aire, upregulate a set of genes, such as keratins (Krt1, 10, 77) or involucrin and form Hassall’s corpuscules^40,41,54. However, the regulatory mechanism(s) guiding this differentiation process remains poorly understood⁵⁵. Our transcriptomic results are consistent with the idea that TLR/MyD88 signaling establishes an expression profile that is associated with the differentiation of mTECs^high into post-Aire mTECs. Notably, TLR9 stimulation not only increased the number of Involucrin⁺ post-Aire mTECs (Supplementary Fig. 3e, f), but also lead to the upregulation of cytokines and chemokines (Il1f6, Lcn2, Cxcl3, and Cxcl5) associated with Hassall´s corpuscles⁴² which attract CD14⁺moDCs. Together with the fact that they serve as a reservoir of a large amount of Aire-dependent TRAs, post-Aire mTECs could hold central position in the mechanism of transfer of mTEC-derived antigens to thymic DCs.

As described above, TLR/MyD88 signaling in mTECs^high drive the expression of chemokines which act on an overlapping set of receptors³² predominantly expressed by CD14⁺moDCs (Cxcr2, Ccr1, Ccr3, Ccr5) and pDCs (Ccr5, Ccr6, and Ccr9). A correlative nature between the frequency of CD14⁺moDC in the thymus of MyD88^ΔTECs and of WT stimulated with CpG, underpins the importance of these chemokines in controlling the migration of these cells into the thymic medulla. However, the deletion of Cxcr2 on hematopoietic cells, the common receptor for Cxcl1, Cxcl2, Cxcl3 and Cxcl5, did not yield any changes in the enrichment of CD14⁺moDC in the thymus (Supplementary Fig. 6h). This observation, in conjunction with previous reports^18,56, allows one to predict that while the ligands of Ccr3 and/or Ccr5 (Ccl3, Ccl4, Ccl5 or Ccl24) likely regulate the entry of CD14⁺moDC into the thymus¹⁹, Cxcl-chemokines may regulate the positioning of these cells in close proximity of post-Aire mTECs. Interestingly, with the decreased frequency of CD14⁺moDCs in the thymus of MyD88^ΔTEC, pDCs were similarly diminished. However, in contrast to CD14⁺moDCs, the number of pDCs did not increase after TLR9 intrathymic stimulation. This is consistent with the fact that the migration of pDCs to the thymus is driven by Ccl25 (ligand of Ccr9 receptor)¹⁴, the expression of which was diminished in MyD88^ΔTEC but was not upregulated in WT mTECs after TLR9 stimulation.

It has been previously documented that specific subtypes of thymic DCs vary in their capacity to acquire antigens from TECs. Notably, while the transfer of MHC molecules from TECs to CD8α⁺cDC1 and Sirpα⁺DCs occurred at the same efficiency¹⁶, the transfer of intracellular GFP was restricted mainly to CD8α⁺cDC1¹⁰. In comparison, our data shows that cytoplasmic TdTOM from Foxn1^CreROSA26^TdTOMATO could to certain extent, be transferred to all major subtypes of thymic DCs. This may be explained by the robustness of the Foxn1^Cre-dependent system where, compared to Aire-GFP model, the production of TdTOM is not restricted only to Aire-expressing mTECs but to the entire thymic TEC population. Importantly, since the CAT of TdTOM after CpG ODN intrathymic injection is increasingly targeted to CD14⁺moDC subpopulation, the efficiency of CAT correlates not only with the broadness of antigen expression but also with the frequency of a given DC subtype in the medulla. On the other hand, since TECs constitute a relatively rare cell population of thymic cells⁵⁷, the amount of antigen, which can be potentially transferred to DCs, is fairly limited. This could explain the fact that even when the entire population of thymic pDCs is not affected by intrathymic TLR9 stimulation, the frequency of TdTOM⁺ pDCs is significantly decreased, due to the increased competition for TdTOM uptake by CD14⁺moDCs.

It has become clear that developing thymocytes encounter self-antigens presented by various types of thymic APC, including mTECs⁴⁷, B-cells⁵⁸, pDCs¹⁴, and cDCs^11,59. Although the generation of thymic Tregs was shown to be dependent on antigen presentation by both mTECs and DCs^4,47, thymic cDCs seem to be particularly important for this process^6,17,60. Along with self-antigen presentation, thymic cDCs express high levels of co-stimulatory molecules CD80/86 as well as CD70 which play a crucial role in promoting thymic Treg development^61,62. Among cDCs, Sirpα⁺DCs are the most efficient in supporting Treg generation^17,20,63. In this context, our data demonstrates that the development of thymic CD25⁺Foxp3⁺ Tregs is boosted by TLR/MyD88 signaling in TECs, which produce a chemokine gradient driving the migration of CD14⁺moDCs into the thymus. We also found that mTEC-intrinsic TLR9/MyD88 signaling increased the cell ratio of Sirpα⁺DCs to Xcr1⁺cDC1, which correlated with an increased production of thymic Tregs. These findings accurately recapitulate the thymic phenotype of Ccr7^–/– mice where the increased ratio of Sirpα⁺DCs to cDC1 correlated with the increased generation of thymic Tregs²⁰. This data, together with the fact that abrogation of MHCII-antigen presentation specifically in DCs, resulted in a reduced number of thymic Tregs in unstimulated¹⁷ as well as in CpG stimulated thymus (Fig. 6h), suggest that TLR/MyDdependent generation of thymic Tregs is mediated by antigen-presentation by DCs.

Our results also show that TLR/MyD88 signalling in mTECs drives the recirculation of mature CD73⁺CD25⁺Foxp3⁺ Tregs into the thymus. Compared to newly generated CD73^– Tregs, their increased number in the TLR9 stimulated thymus was not dependent on MHCII presentation by DCs. Together, with the fact that recirculation of CD73⁺ Tregs was not abrogated in MyD88^ΔTECs mice, suggests that Ccl20, the ligand for Ccr6, which is highly expressed by recirculating Tregs⁶⁴ regulates the increased recirculation of CD73⁺CD25⁺Foxp3⁺ Tregs into the thymus after TLR9 intrathymic stimulation (Figs. 2d and e).

Altogether, our model proposes that TLR/MyD88 signaling in mTECs regulates the generation of Tregs. The mechanism involves TLR-induced chemokine production and subsequent chemotactic recruitment of CD14⁺moDC to the thymic medulla, which predicates the developmental output of Tregs. Although this study explores only TLR9 signaling in mTECs, questions surrounding the nature of potential thymic-derived endogenous ligands for TLR/MyD88 signals in mTECs remains enigmatic and warrant further study.

Methods

Mice

A majority of the mice used in this study were of C57BL/6J genetic background and housed in the animal facility at the Institute of Molecular Genetics of the ASCR v.v.i. under SPF conditions. Mice were fed with irradiated standard rodent high energy breeding diet (Altromin IRR) and given reverse osmosis filtered and chlorinated water ad libitum. Light were adjusted to a 12 h/12 h light/dark cycle; temperature and relative humidity were maintained at 22 ± 1°C and 55 ± 5%, respectively. Experimental protocols were approved by the ethical committee of the Institute of Molecular Genetics and by the ethical committee of the Czech Academy of Science. Aire^–/– (BS2-Aire^tmDoi/J, stock# )², Foxn1^Cre (B6(Cg)-Foxn1^tm3(cre)Nrm/J, stock# )²⁹, MyD88^fl/fl (BP2(SJL)-Myd88^tm1Defr/J, stock# ), MyD88^–/– (BP2(SJL)-Myd88^tmDefr/J,, stock# )³⁰, Rag1^–/– (BS7-Rag1^tm1Mom/J, stock# )⁶⁵, Ly (eunic-brussels.eu-Ptprc^aPepc^b/BoyJ, stock# )⁶⁶, Cxcr2^fl/fl (C57BL/6-Cxcr2^tm1Rmra/J, stock# )⁶⁷, H2-Ab1^fl/fl (B×1-H2-Ab1^tm1Koni/J, stock# )⁶⁸, and Itgax^Cre (eunic-brussels.eu-Tg(Itgax-cre)Reiz/J, stock# )⁶⁹ mice were purchased from Jackson Laboratories. Rosa26^TdTOMATO (B6;S6-Gt(ROSA)26^{Sortm14(CAG-tdTomato)Hze}/J, stock# )⁷⁰ and Vav1^Cre (eunic-brussels.eu-Commd10^{Tg(Vav1-icre)A2Kio}/J, stock# )⁷¹ were kindly provided by V. Kořínek (Institute of Molecular Genetics of the ASCR, Prague, Czech Republic). Aire-HCO (Balb/c)⁴ were provided by L. Klein. Cd3ε^–/–⁷², RIP-OVA⁷³, OT-I⁺Rag2^–/–⁷⁴ (all C57BL/6J) were provided by O. Štěpánek. OT-II (eunic-brussels.eu-Tg(TcraTcrb)Cbn/J, stock# )⁷⁵ mice were kindly provided by T. Brdička (Institute of Molecular Genetics of the ASCR, Prague, Czech Republic). C57BL/6J GF and control C57BL/6J SPF mice were kindly provided by M. Schwarzer (Institute of Microbiology of the ASCR, Nový Hrádek, Czech Republic. Both GF and control SPF mice were subject to the SSNIFF V diet. Thymic cell populations were isolated from 3–6-week-old mice with the exception of newborn mice (4 days old) used in Supplementary Fig. 7b. For the purpose of BM chimera experiments, 5–6-week-old mice were irradiated and analysed between 11 and 13 weeks of age. Comparative analysis used age-matched cohorts regardless of sex and caging. Where possible, littermates were used as the controls. For the purpose of tissue isolation, mice were euthanized by cervical dislocation.

Tissue preparation and cell isolation

Thymic antigen presenting cells, TECs and DCs, were isolated as follows. Thymus was minced into small pieces and treated with Dispase II (Gibco), dissolved in RPMI at concentration mg ml⁻¹. Tissue was homogenized by pipetting and after 10 min of incubation (37°C), the supernatant was collected and the reaction was stopped by adding 3% FSC and 2 nM EDTA. The process was repeated until all thymic fragments were digested. For detail description see⁷⁶. For thymic epithelial cells isolation, the whole thymic cell suspension was depleted of CD45⁺ cells by CD45 microbeads staining (Miltenyi biotec). Thymic dendritic cells were isolated using MACS enrichment for CD11c⁺ cells through staining with biotinylated CD11c antibody, followed by Ultrapure Anti-Biotin microbeads staining (Miltenyi biotec). For isolation of T cell, thymus, peripheral lymph nodes (pLN), mesenteric lymph nodes (mLN) or spleen were mechanically mashed through 40 μm Cell strainer (Biologix) and cell suspensions were passed through 50 μm filters (Sysmex). The resulting cell suspension was spun down (4 °C, g, 10 min) and erythrocytes were removed using ACK lysis buffer.

Flow cytometry analysis and cell sorting

Flow cytometry (FACS) analysis and cell sorting were performed using BD LSR II and BD Influx (BD Bioscience) cytometers, respectively. For surface staining, cells were incubated for 20–30 min at 4 °C with the indicated fluorochrome- or biotin-conjugated antibodies. Where necessary, cells were further incubated with streptavidin conjugates for 15 min. Dead cells were excluded using Hoechst (Sigma) or viability dye eFlour or (eBioscience). For the intracellular staining of Aire and Foxp3, the cells were first stained for the targeted surface molecules, fixed, and permeabilized for 30 min at room temperature (RT) using the Foxp3/Transcription Factor Staining Buffer Set (eBioscience), then stained for 30 min at RT with fluorochrome-conjugated antibodies. FlowJO V10 software (Treestar) and BD FACSDiva™ Software v for BD™ LSR II (with HTS Option) was used for FACS data analysis including tSNE analysis shown in Fig. 5e. A complete inventory of staining reagents is listed in Supplementary Data 6.

Imaging flow cytometry

Imaging flow cytometry was performed at the Center for Advanced Preclinical Imaging (CAPI) with the use of AMNIS ImageStream X MkII (AMNIS). DCs isolated from Foxn1^CreROSA26^TdTOMATO mice were stained for the surface markers MHCII and CD11c. Dead cells were excluded by Hoechst staining and bright field analysis. Cells were recorded using 40x magnification. Data was analyzed with Ideas software (AMNIS). A complete list of staining reagents can be found in Supplementary Data 6.

In vitro TLRs and cytokines stimulation assays

mTECs^high were gated as EpCAM⁺CD11c^–Ly51^–MHCII^highCD80^high and sorted into RPMI media (Sigma) containing 10% FSC and 1% Penicillin/Streptomycin (Gibco). Cells were then cultured in a flat-well plate in μL of 10% FSC RPMI with Penicillin/Streptomycin in the presence of Endotoxin-free TLR ligands (InvivoGen) or recombinant mouse cytokines: TLR9 ligand-CpG ODN (ODN ) (5 μM), TLR4 ligand-LPS (1 μg/ml), Il-1β (10 ng/ml), Il (10 ng/ml) (both ImmunoTools) and Il (10 ng/ml) (Biolegend). After 24 h, the supernatant was removed and the cells were resuspended in RNA-lysis buffer. Subsequently, RNA isolation was performed.

In vivo TLR stimulation

For intrathymic injections, mice were anesthetized by i.p. injection of Zoletil (Tiletamine (50 mg/ml) and Zolazepam (50 mg/ml), Virbac) dissolved in PBS at a dose of 50 mg/kg and 10–20 μl of μM CpG ODN (ODN , InvivoGen) or PBS was injected using an insulin syringe (29G) directly into the first intercostal space from the manubrium ~2 mm left of the sternum and 4 mm in depth. The angle of injection was from 25 to 30° relative to the sternum⁷⁷. For systemic TLR9 stimulation, mice were injected by CpG ODN (ODN , InvivoGen) ( μM) or PBS at day 0 and day 1 into the peritoneum. Mice were then maintained under SPF conditions and euthanized at the indicated time point of an experiment.

Immunofluorescent analysis of thymic cryosections

The thymus was fixed overnight in 4% paraformaldehyde (Sigma) at 4 °C, washed three times in PBS, incubated overnight in 30% sucrose at 4 °C, and finally embedded in OCT compound (VWR). Cryoblocks were cut at 8 μm and blocked with PBS containing 5% BSA (w/v) and % Triton X for 1 hour at room temperature. Samples were incubated overnight at 4 °C with the following primary antibodies: anti-keratin 14, Sirpα, and CD11c-biotin (Fig. 3d) or anti-Involucrin and anti-EpCAM-APC (Supplementary Fig. 3b). The samples were stained with secondary reagents, Goat anti-rat AF, goat anti-rabbit AF and streptavidin FITC or goat anti-rabbit AF for one hour at RT. Sections stained only with secondary reagents were used as negative controls. 4′,6-diaminophenylindole (DAPI) was used to visualize cell nuclei. Stained sections were mounted in Vectashield medium (Vector Laboratories) and imaged using a Dragonfly (Andor)—spinning disk confocal microscope with the immersion objective HC PL APO 20×/ A complete list of staining reagents can be found in Supplementary Data 6. Z-stacks were composed using ImageJ and deconvolution was done by Huygens Professional. CD11c⁺Sirpα⁺ double positive cells were counted in multiple μmxμm areas in keratin rich (medulla) and keratin negative (cortex) region. Counting was done as a blind experiment by three different investigators. Involucrin⁺EpCAM⁺ double positive cells were counted as number of cells per thymic medullary region (determined by DAPI staining).

Gene expression analysis by qRT-PCR

Total RNA from FACS-sorted cells was extracted using an RNeasy Plus Micro Kit (Qiagen) and reverse transcribed using RevertAid (ThermoFisher) transcriptase and random hexamers (ThermoFisher). Quantitative RT PCR (qRT PCR) was performed using the LightCycler SYBR Green I Master mix (Roche) on a LightCycler II (Roche). Each sample was tested in duplicate. Threshold cycles were calculated using LightCycler software. Gene expression was calculated by the relative quantification model⁷⁸ using the mRNA levels of the housekeeping gene, Casc3, as a control. Primers were designed using Primer-BLAST (NCBI, NIH). Primers sequences are listed in Supplementary Data 6.

Bone marrow chimera generation

Bone marrow cells were isolated from the femur and tibia of Ly mice (CD⁺) and subsequently depleted of erythrocytes using ACK lysis buffer. Recipient mice (Foxn1^CreROSA26^TdTOMATO, CD⁺) were irradiated with 6 Gy and reconstituted with 2 × 10⁶ donor BM cells. These mice were maintained on water supplemented with gentamycin (1 mg/ml) for 10 days. Three weeks after irradiation, the frequency of blood cell reconstitution was measured by FACS using anti-CD and CD antibodies. If the reconstitution was higher than 80%, mice were euthanized 6 weeks after transfer and subjected to further analysis.

RNA-sequencing and analysis

mTECs were sorted according to the protocol described above and RNA was extracted using a RNeasy Plus Micro Kit (Qiagen). cDNA synthesis, ligation of sequencing adaptors and indexes, ribosomal cDNA depletion, final PCR amplification and product purification were prepared with a SMARTer® Stranded Total RNA-Seq – Pico Input Mammalian library preparation kit v2 (Takara). Library size distribution was evaluated on a Agilent Bioanalyzer using the High Sensitivity DNA Kit (Agilent). Libraries were sequenced on a Illumina NextSeq® instrument using a 76 bp single-end high-output configuration resulting in ~30 million reads per sample. Read quality was assessed by FastQC (). Subsequent read processing including removing sequencing adaptors (Trim Galore!, version ), mapping to the reference genome (GRCm38 (Ensembl assembly version 91)) with HISAT2 (), and quantifying expression at the genetic level (featureCounts) was done via the SciLifeLab/NGI-RNAseq pipeline [eunic-brussels.eu]. Final per gene read counts served as an input for differential expression analysis using a DESeq2 R Bioconductor (). Prior to this analysis, genes that were not expressed in at least two samples were discarded. Genes exhibiting a minimal absolute log2-fold change value of 1 and a statistical significance (adjusted p-value < ) between conditions were considered as differentially expressed for subsequent interpretation and visualization. All figures (volcano plots, etc.) were generated using basic R graphical functions. The raw sequencing data were deposited at the ArrayExpress database under accession numbers E-MTAB (for Fig. 2a, b) and E-MTAB (for Fig. 2d, e).

Single-cell RNA sequencing

DCs were sorted from Foxn1^CreROSA26^TdTOMATO as Gr-1^–CD11c⁺TdTOM⁺ (described in detail in Supplementary Fig. 3a and Fig. 4c). Two independent samples (Sample 1 and 2) were used for further analysis. A single-cell library was prepared by Illumina/Bio-Rad single-cell RNA-seq system with a SureCell WTA 3’ Library Prep Kit according to the manufacturer’s instructions. Total cell concentration and viability was ascertained using a TC20 Automated Cell Counter (Bio-Rad). A ddSEQ Single-Cell Isolator (Bio-Rad) was used to co-encapsulate single cells with barcodes and enzyme solutions for cDNA synthesis. Nextera SureCell transposome solution was used for cDNA fragmentation and ligation of sequencing indexes, followed by PCR amplification and short fragment removal. Finally, library fragment length distribution and concentration were analyzed on a Agilent Bioanalyzer using a High Sensitivity DNA Kit (Agilent). The resulting libraries were sequenced using a 68/75 paired-end configuration on a Illumina NextSeq® instrument resulting in ~73 million reads per sample.

Single-cell RNA sequencing analysis

The quality of reads was assessed by FastQC. Cell identification was accomplished with cell barcodes and low-expression cells filtering using UMI-tools⁷⁹. The analysis identified cells in Sample 1 and cells in Sample 2. Reads assigned to the selected cells were mapped to the GRCm38 genome assembly (Ensembl version 91) with HISAT2 (). Gene expression was quantified using, featureCounts () after deduplication of per-gene assigned read counts by UMIs with UMI-tools. De-duplicated per-gene read counts were imported into R for exploration and statistical analysis using a Seurat⁸⁰ package (version ). Counts were normalized according to total expression, multiplied by a scale factor (10,), and log-transformed. For cell cluster identification and visualization, gene expression values were also scaled according to highly variable genes after controlling for unwanted variation generated by sample identity. Cell clusters were identified based on t-SNE of the first six principal components of PCA using Seurat’s method, FindClusters, with a original Louvain algorithm and resolution parameter value of To find cluster marker genes, Seurat’s method, FindAllMarkers, along with a likelihood ratio test assuming an underlying negative binomial distribution suitable for UMI datasets was used. Only genes exhibiting a significant (adjusted p-value < ) minimal average absolute log2-fold change of 1 between each of the clusters and the rest of the dataset were considered as differentially expressed. For t-SNE expression plots, normalized count data were used. Heatmaps of gene expression per cluster were generated based on gene z-score scaled raw counts. The raw sequencing data was deposited at the ArrayExpress database under accession number E-MTAB

In vitro antigen presenting assay

For the purpose of antigen presentation assay CD14⁺moDCs were gated as CD11c⁺MHCII⁺B^–Xcr1^–Cx3cr1⁺CD14⁺ and FACS sorted from Aire-HCO mice into DMEM high-glucose medium (Sigma) supplemented with 10% FCS and 1% Penicillin-Streptomycin (Gibco) and cultivated in a 96 well plate together with the A5 hybridoma cell line (HA-specific CD4 T cell hybridoma cells carrying a GFP-NFAT reporter) at a ratio (10 of CD14⁺moDC: 50 of A5 cells). As a positive control, CD14⁺moDCs were pulsed with HA peptide (; customized by Thermofisher) at a concentration of 1 μg/ml. After 20 h, the level of GFP expression by A5 hybridomas was analyzed by flow cytometry.

Induction of T cell transfer colitis and histological analysis

FACS-sorted 5×10⁵ TCRβ⁺CD4⁺CD45RB^highCD25^– or complete TCRβ⁺CD4⁺ were transferred by i.p. injection into Rag1^–/– recipient mice ( weeks old). The weight of mice was recorded weekly to monitor the progress of colitis. Mice were euthanized 8 weeks after transfer⁵¹. Spleens and colons of the animals were weighed and the length of the colon was measured. For histological analysis PBS washed colons were fixed in 4% paraformaldehyde (Sigma) and embedded into paraffin. Tissue sections were cut into 5μm thin slices, deparaffinized, and stained with hematoxylin and eosin (H&E).

In vitro Tregs suppression assay

BM-derived DCs (BMDCs) were prepared as follows. BM cells were flushed from femur and tibia of WT C57BL/6J mice and cultured in RPMI media (Sigma) containing 10% FSC and 1% Penicillin/Streptomycin (Gibco) supplemented with GM-CSF (5 ng/ml). Fresh media containing GM-CSF was added at day 3 and 5 of cultivation. After 7 days, BMDCs was pulsed with OVA cognate peptide (irrelevant OVA – peptide was used as control) (InvivoGen) at a concentration of 1 μg/ml and co-cultivated with OVA-specific OT-II T cells and Tregs (10 BMDCs: 50 OT-II T cells: 50 Tregs). OT-II T cells were isolated from OT-II⁺Rag1^−/− mice as MACS-enriched CD4⁺ T cells (CD4⁺ T Cell Isolation Kit, Miltenyi biotec). CD4⁺ conventional T cells (Tconv) were used as a negative control. Tregs were isolated from LNs (pLN and mLN) of WT (MyD88^fl/fl) and MyD88^ΔTECs mice using subsequent Auto-MACS (Miltenyi biotec) procedure. CD4-enriched T cells (CD4⁺ T Cell Isolation Kit, Miltenyi biotec) were stained by anti-CD25 biotin conjugated antibody and CD4⁺CD25⁺ Tregs were isolated using Anti-Biotin MicroBeads (Miltenyi biotec). Tconv cells were prepared using Auto-MACS as CD4⁺CD25^– cells. After 3 days of co-cultivation, cells were stained with anti-Vβ5 and anti-Vα2 antibodies to distinguish OT-II⁺ T cells. Proliferation was measured by FACS using CPD staining.

In vivo model of autoimmune diabetes

Cd3ε^–/–RIP-OVA mice (6–8 weeks old) were intravenously injected by MACS enriched CD8⁺ T cells (5 × 10⁵ cells per mouse) isolated from lymph nodes and spleen of Rip-OVA Ly (CD⁺) mice at day 8. After 7 days (day 1) Cd3ε^–/–RIP-OVA mice were intravenously injected, FACS sorted CD4⁺CD25⁺ Tregs were isolated from LNs (mLN and pLN) of WT (MyD88^fl/fl), MyD88^ΔTECs mice (3×10⁵ cells per mouse), OT-I (OT-I⁺Rag2^–/–, cells per mouse), and OT-II cells (OT-II⁺Rag1^–/–, 1 × 10⁴ cells per mouse). BMDCs (generated as described previously, 10 days of culture, media refreshment at day 4 and 7) were pulsed with OVA peptides (OVA –, 2 mM and OVA , μM, InvivoGen) in the presence of LPS ( μg/ml, InvivoGen) for 3 h. In all, 1 × 10⁶ of antigen-stimulated DCs were used for injection (at day 0). Glucose levels were monitored on a daily basis (between day 5 and 14) using test strips (Diabur-Test , Roche or GLUKOPHAN, Erba Lachema, Czech Republic). The animal was considered to have developed autoimmunity when the concentration of glucose in the urine reached ≥10 mmol/l. At day 14, mice were euthanized and the frequency of splenic KLGR1⁺ OT-1 T cells was measured by flow cytometry.

Statistical analysis

The statistical tests used to analyze the data are indicated in figure legends. Graph construction and statistical analysis were performed using Prism software (GraphPad). Statistical analysis of RNAseq and scRNAseq data is indicated in the corresponding method section.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

The authors declare that all data supporting the findings of this study are available within the article and its supplementary information files or from the corresponding author upon reasonable request. The source data underlying Fig. 1c, f, 2c, f, g, 3a–c, e, 4d, f, 5d, f, g, 6a, c, e–i, 7b, d–h and Supplementary Figs. 2b–d, 3a, b, d, e, 4c, 5e, f, 6c, e, h, 7a–d, f, h and 8b, c, e are provided as a Source Data file. The raw RNA sequencing data are deposited at the ArrayExpress database [eunic-brussels.eu] under accession numbers E-MTAB (Fig. 2a, b), E-MTAB (Fig. 2d, e) and E-MTAB (Fig. 5a–c).

Change history

References

Klein, L., Kyewski, B., Allen, P. M. & Hogquist, K. A. Positive and negative selection of the T cell repertoire: what thymocytes see (and don’t see). Nat. Rev. Immunol.14, – ().
CASPubMedPubMed Central Google Scholar
Anderson, M. S. et al. Projection of an immunological self shadow within the thymus by the aire protein. Science, – ().
ADSCASPubMed Google Scholar
Liston, A., Lesage, S., Wilson, J., Peltonen, L. & Goodnow, C. C. Aire regulates negative selection of organ-specific T cells. Nat. Immunol.4, – ().
CASPubMed

Monocyte biology conserved across species: Functional insights from cattle

Introduction

With their high functional plasticity (1, 2), monocytes are a central component of the mononuclear phagocyte system (MPS). Although their delineation from bona fide macrophages and bona fide dendritic cells has proved challenging, especially in tissues, monocytes and monocyte-derived cells are now fully appreciated as a separate lineage (2&#x;4). In blood of humans and cattle, monocytes can be subdivided into at least two different subsets based on the expression of CD14 and CD16 (5&#x;7): classical monocytes (cM; CD14^highCD16^-) and nonclassical monocytes (ncM; CD14^-/dimCD16^high). In mice, analogous subsets can be defined by Ly6C expression (8). A third intermediate monocyte subset (intM; CD14^highCD16^+/high) is less well defined and has been shown to transcriptionally resemble ncM in both humans and cattle (5, 6, 9). In these two species, cM are described as the dominant subset in peripheral blood comprising about 80% of all monocytes, while ncM and intM comprise only small fractions (about 10% each) (2, 7). In mice, however, ncM (Ly6C^-) are reported to be as frequent as cM (Ly6C⁺) in peripheral blood (8).

Classical monocytes are known for their pro-inflammatory function especially in bacterial infections (10), however the role of intM and ncM is less well described. Nonclassical monocytes are generally viewed as anti-inflammatory and vasoprotective (11), as they were found to crawl along vascular endothelium (8, 12) and sustain vascular integrity by orchestrating endothelial renewal (13). The prominent transcription of genes for endothelial adhesion in bovine ncM (7, 9) suggests a similar role in cattle. In response to TLR7 stimulation, murine ncM have been shown to recruit neutrophils to the endothelium and to clear neutrophil-induced focal necrosis (13). Also in humans, ncM were shown to be specialized in sensing nucleic acids via TLR7 and TLR8 (14) and are proposed to function in antiviral immunity (15). Murine ncM have furthermore been described as biased progenitors of wound healing macrophages (16).

Monocytes in general are known to be capable of antigen presentation to T cells, however whether they are as potent as dendritic cells remains controversial (1). Notably, TLR7 stimulation, but not TLR3 or TLR4 stimulation, has been shown to promote cross-presenting abilities in murine Ly6C⁺ cM (17). Bovine monocytes, particularly ncM, were reported to induce allogeneic T-cell responses in vitro (18).

We have previously reported pronounced transcriptomic differences between bovine cM and ncM, as determined by bulk RNA-seq and principal component analysis (9). The present study explores these transcriptomic differences in greater detail: we performed in-depth gene-by-gene analysis of bulk- and single-cell transcriptomes, as well as analyses of TLR responsiveness and metabolic activity. Taken together, our data indicate subset-specific functions in acute inflammation, antibacterial and antiviral responses, as well as in T-cell modulation, resolution of inflammation, and tissue repair. Furthermore, the unsupervised clustering of our single-cell RNA-seq data confirms the CD14/CDbased subset definition, but also supports continuous differentiation of bovine monocyte subsets &#x; yet another feature presumably shared with their human counterparts (19).

Materials and methods

Isolation of bovine PBMC

Blood of female cattle (Bos taurus; various breeds; aged months to 9 years) was collected at the Clinic for Ruminants (Vetsuisse Faculty, University of Bern, Switzerland) or at the animal facility of the Institute of Virology and Immunology (Mittelhäusern, Switzerland) by puncturing the jugular vein, using citrate-based Alsever&#x;s solution ( mM of C₆H₁₂O₆, mM of Na₃C₆H₅O₇·2H₂O, mM of NaCl, and 43 mM of C₆H₈O₇, pH ) as an anticoagulant. Blood sampling was performed in compliance with the Swiss animal protection law and approved by the cantonal veterinary authority (license numbers BE/15, BE/17, and BE/17).

For peripheral blood mononuclear cell (PBMC) isolation, blood was centrifuged at x g for 20 min (20°C), the buffy coat was collected, diluted with PBS to a ratio of 1 to 1 (room temperature), and layered onto lymphocyte separation medium ( g/mL; GE Healthcare). After centrifugation ( x g, 25 min, 20°C), PBMC were collected and washed twice ( x g, 8 min, 4°C) with cold PBS containing 1 mM UltraPure&#x; EDTA (ThermoFisher). In order to remove platelets, a final washing step was performed at x g (8 min, 4°C).

Phenotyping of monocyte subsets by flow cytometry

Phenotyping of bovine monocyte subsets was performed with freshly isolated PBMC in well. U-bottom microtiter plates (1 x 10⁷ cells per sample). Antibodies used for the two-step five-color stainings are shown in Table 1. Incubations were performed for 20 min at 4°C. Washing steps between incubations ( x g, 4 min, 4°C) were done with Cell Wash (BD Biosciences). Prior to staining, PBMC were incubated with bovine IgG in order to block Fc receptors (50 µg/mL; Bethyl laboratories). For detection of dead cells, Live/Dead Near-IR stain (ThermoFisher) was included in the last incubation step. Compensation was calculated by FACSDiva software using single-stained samples. For each marker to be examined on monocyte subsets, a fluorescence-minus-one (FMO) control was included. Samples were acquired with a FACSCanto II flow cytometer (BD Biosciences) equipped with three lasers (, , and nm). At least 5 x 10⁵ cells were recorded in the large-cell gate.

Table 1 Antibodies used for flow-cytometric phenotyping of bovine monocyte subsets.

Fluorescence-activated cell sorting and bulk RNA sequencing

Bulk RNA sequencing data of sorted monocyte subsets are derived from previous experiments, described in Talker et al. (9). Experimental procedures are therefore described only briefly and the reader is referred to our previous publication for more details. In order to sort bovine monocyte subsets for Illumina sequencing, a two-step staining was performed with 3 x 10⁸ freshly isolated PBMC. Classical monocytes (cM) were sorted as CDa^highCD14^highCD16^-, intM as CDa^highCD14^highCD16^high and ncM as CDa^highCD14^-/dimCD16^high using a FACS Aria (BD Biosciences). All sorted subsets had a purity of at least 97%. Per subset, at least 1 x 10⁵ sorted cells were frozen in TRIzol (ThermoFisher) for later RNA extraction (Nucleospin RNA kit, Macherey Nagel). High-quality RNA (approximately ng; RIN> 8) was used for nondirectional paired-end mRNA library preparation (TruSeq Sample Preparation Kit; Illumina).

Sequencing was performed on the Illumina HiSeq platform using bp single-end sequencing, yielding between and million read pairs per sample. Reads were mapped to the Bos taurus reference genome (UMD) with Hisat2 v. , and FeatureCounts from Subread v. was used to count the number of reads overlapping with each gene, as specified in the Ensembl annotation (release 91). Raw counts of the sequencing data previously published in Talker et al., (9) were re-analyzed with the Bioconductor package DESeq2 (20), including only data for the three monocyte subsets and considering the factor animal in the design formula. Raw counts were normalized to account for differences in sequencing depth between samples. Gene length was not considered. Lists of differentially expressed genes were obtained by performing pairwise comparisons with DESeq2 (adjusted p-value < ). Gene lists were manually screened for genes of interest using the human gene database GeneCards^® and literature research via PubMed^®. Principal component analysis was performed with normalized and rlog-transformed counts of the most variable genes across monocyte samples. Heatmaps were prepared following log2 transformation of normalized counts. Prior to log2 transformation, a pseudocount of 1 was added to the values to avoid zeros. All analyses were performed using R version

For gene set enrichment analysis (GSEA) we used genes ranked based on the DESeq2 output (cM vs. ncM, filtered for p_adj < ) according to the stat value representing the Wald statistics. We employed the GSEA software from UC San Diego and Broad Institute (21, 22) with the C5:GO:BP gene sets (Gene Ontology, biological process, file eunic-brussels.eu) that are integrated into the software (MSigDB v) (22, 23). The following parameters were used: gene set size ; scoring scheme weighted , normalization meandiv , mode mean of probes , number of permutations .

The bulk-RNA-seq datasets are available in the European Nucleotide Archive (eunic-brussels.eu) under the accession number PRJEB

Phosphoflow cytometry

The following TLR ligands were used at a final concentration of 10 μg/mL to assess TLR responsiveness of bovine monocyte subsets: Pam2CSK4 [C₆₅HN₁₀O₁₂S &#x; 3TFA] (InvivoGen), high-molecular-weight (HMW) polyinosinic-polycytidylic acid [Poly(I:C)] (InvivoGen), LPS from E. coli strain K12 [LPS-EK] (InvivoGen), Gardiquimod (Sigma-Aldrich), Resiquimod (Sigma-Aldrich).

Stimulation with TLR ligands and staining for phosphorylated p38 MAPK was performed as reported previously (24). Prior to stimulation, cell surface staining was performed. Following Fc-receptor blocking with purified bovine IgG (50 µg/mL; Bethyl laboratories), defrosted and CD3-depleted bovine PBMC were stained with anti-CDa (CC, IgG2b), anti-CD14 (CAM36A, IgG1), and anti-CD16 (KD1, IgG2a), followed by incubation with anti-IgG2b-AF (Molecular Probes), anti-IgG1-biotin (Southern Biotech) and anti-IgG2a-PE (Southern Biotech). In a fourth step, ChromPure mouse IgG (Jackson ImmunoResearch) was added together with Streptavidin-BV (BD Biosciences) and Live/Dead&#x; Fixable Near-IR stain (Thermo Fisher Scientific). Stained cells were incubated for 15 minutes with respective TLR ligands in PBS or with PBS alone (waterbath at 37°C). Immediately after this incubation period, cells were fixed with BD Cytofix/Cytoperm&#x; fixation buffer (BD Biosciences; 12 min at 37°C) and thereafter stained with anti-p38 MAPK-AF (BD Phosflow; 30 min at 37°C). Samples were acquired with a FACSCanto II flow cytometer (BD Biosciences) equipped with three lasers (, , and nm). At least × 10⁶ cells were recorded in the large-cell gate. Compensation was calculated by FACSDiva software using single-stained samples.

Extracellular flux analysis

For metabolic assays using a Seahorse Extracellular Flux Analyzer (denominated Seahorse assays ; Agilent Technologies Inc.), bovine monocyte subsets were FACS-sorted from magnetically enriched CDa⁺ PBMC. Combined staining for magnetic selection and FACS was performed in 50 mL centrifugation tubes and included five incubation steps, each carried out at 4°C for 20 min. Staining, and washing ( x g, 8 min, 4°C) in-between each incubation step, was performed in PBS containing 1 mM EDTA and 5% (v/v) heat-inactivated FBS (GIBCO, Life Technologies). In a first step, freshly isolated PBMC (4 x 10⁸) were incubated with bovine IgG (50 µg/mL; Bethyl laboratories) to block Fc receptors. This was followed by incubation with the primary antibodies anti-CDa (CC, IgG2b), anti-CD14 (CAM36A, IgG1), and anti-CD16 (KD1, IgG2a) and the secondary antibodies anti-IgG1-AF (Molecular Probes) and anti-IgG2a-PE (Southern Biotech). In a fourth step, anti-mouse IgG magnetic beads (Miltenyi Biotec) were added, and cells were loaded onto two LS columns (Miltenyi Biotec) for magnetic enrichment of CDa expressing cells. In a fifth step, anti-IgG2b-AF (Molecular Probes) was added, resulting in a dim staining of CDa for FACS. Enriched monocytes were sorted on a MoFlo Astrios EQ cell sorter (Beckman Coulter) equipped with five lasers at the Flow Cytometry and Cell Sorting Facility (FCCS) of the University of Bern. Purity of subsets was confirmed by re-analysis of samples and was shown to be at least 98%. Following sorting, cells were resuspended in pH-optimized Seahorse assay medium (Agilent Seahorse XF DMEM Medium supplemented with 10 mM Agilent Seahorse XF Glucose, 1mM Agilent Seahorse XF Pyruvate, and 2mM Agilent Seahorse XF Glutamine) and 5 x 10⁵ cells (ex vivo) or 1 x 10⁵ cells (after 6-day culture) were seeded in duplicates or triplicates into an 8-well Seahorse plate (Agilent Seahorse XFp FluxPak). Two wells served as background controls. During manual cell counting, viability of sorted subsets was confirmed to be above 95% by trypan blue staining. Cells not immediately used for Seahorse assays, were incubated for six days at 37°C (5% CO₂) to allow for differentiation into monocyte-derived macrophages. Specifically, remaining cells were seeded into a well plate with 1 x 10⁶ cells per well in 2 mL culture medium consisting of DMEM-GlutaMAX&#x; (Gibco, eunic-brussels.eu: ), supplemented with penicillin (50 I.U./mL), streptomycin (50 µg/mL), 10% heat-inactivated FBS, and 1% M-CSF (produced in house; titrated to optimize viability of monocytes in culture). After six days of incubation, cells were harvested, resuspended in Seahorse assay medium, counted, and processed as described above. Viability of harvested cells was above 90% as assessed microscopically by the trypan blue exclusion test.

Seahorse preparations and measurements were performed according to the manufacturer&#x;s instructions. For measurements with the Agilent Seahorse XFp Analyzer, the standard settings of the XFp Real-Time ATP Rate Assay were used. Following basal measurements, oligomycin (final concentration 9 µM), an inhibitor of ATP synthase (complex V of the electron transport chain), was added. After another measurement phase, a mastermix of rotenone (final concentration 8 µM), an inhibitor of complex I, and antimycin A (final concentration 8 µM), an inhibitor of complex III, was added. Inhibitors of the electron transport chain were purchased from Sigma-Aldrich (cat. no. O, R, A) and titrated on bovine cM to determine optimal concentrations. For establishment of the procedure, Agilent Seahorse XFp Real-Time ATP Rate Assay Kit (cat. no. ) was used according to manufacturer&#x;s instructions, but with an increased final concentration (6 µM) of the provided oligomycin, according to prior titration on bovine cM. Data were analyzed using the Seahorse Wave Desktop Software version (Agilent).

Single-cell RNA sequencing (10x Genomics)

For scRNA-seq, PBMC from two cows ( months old; #CH and #CH) were isolated as described above. Cell counting and viability assessments were carried out microscopically using the trypan blue exclusion test. Thereafter, cells were delivered to the Next Generation Sequencing Platform at the University of Bern and processed as follows: GEM generation & barcoding, reverse transcription, cDNA amplification and 3&#x; gene expression library generation steps were all performed according to the Chromium Single Cell 3&#x; Reagent Kits v3 User Guide (10x Genomics CG Rev B) with all stipulated 10x Genomics reagents. Generally, µL of each cell suspension ( cells/µL) and µL of nuclease-free water were used for a targeted cell recovery of 10&#x; cells. GEM generation was followed by a GEM-reverse transcription incubation, a clean-up step and 11 cycles of cDNA amplification. The resulting cDNA was evaluated for quantity and quality using a Thermo Fisher Scientific Qubit fluorometer with the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific, Q) and an Advanced Analytical Fragment Analyzer System using a Fragment Analyzer NGS Fragment Kit (Agilent, DNF), respectively. Thereafter, 3&#x; gene expression libraries were constructed using a sample index PCR step of 16 cycles. The generated cDNA libraries were tested for quantity and quality using fluorometry and capillary electrophoresis as described above. The cDNA libraries were pooled and sequenced with a loading concentration of , paired end and single indexed, on an illumina NovaSeq sequencer using a NovaSeq SP Reagent Kit v1 ( cycles; illumina, ). The read set-up was as follows: read 1: 28 cycles, i7 index: 8 cycles, i5: 0 cycles and read 2: 91 cycles. The quality of the sequencing runs was assessed using illumina Sequencing Analysis Viewer (illumina version ) and all base call files were demultiplexed and converted into FASTQ files using illumina bcl2fastq conversion software v More than 36&#x; reads/cell were generated for each sample.

Mapping and counting of UMIs was performed using Cell Ranger (version , 10x Genomics) with the reference genome ARS-UCD from Ensembl to build the necessary index files. Subsequent analysis was performed in R (version ) (25). The Scater package (version ) (26) was used to assess the proportion of ribosomal and mitochondrial genes as well as the number of detected genes. Cells were considered as outliers and filtered out if the value of the proportion of expressed mitochondrial genes or the number of detected genes deviated more than three median absolute deviations from the median across all cells. Additionally, all ribosomal genes were removed. After quality control, the sample from cow #CH retained cells and the sample from cow #CH retained cells. Normalization between samples was done with the deconvolution method of Lun et al. (27) using the package Scran (version ) (28). Samples were integrated with the FindIntegrationAnchors function of the package Seurat (version ) based on the first 30 principal components (PCs) (29). Graph-based clustering was done with the FindNeighbors and FindClusters functions of the Seurat package using the first 35 PCs from the dimensionality reduction step. The Clustree package (version ) (30) was used to determine the resolution () resulting in clustering concurring with the presumed cell types. Clusters were annotated based on marker genes that were identified with the FindAllMarkers function of Seurat. Cells from clusters identified as monocytic cells (c4, c10, c14, and c18) were extracted and re-clustered in an identical fashion as above, but with a resolution of The scRNA-seq datasets are available in the European Nucleotide Archive (eunic-brussels.eu) under the accession number PRJEB

The R package Monocle3 (31&#x;33) was used to do a trajectory analysis on the clusters of the re-clustered cells. UMAP dimensionality reduction method was used and the Louvain clustering method using nearest neighbors (function cluster_cells of Monocle3). The function graph_test of Monocle3 was used to find genes that are differentially expressed across a single-cell trajectory. Default parameters were used unless stated otherwise.

Ethics Statement

The animal experiments were performed in compliance with the Swiss animal protection law (TSchG SR ; TSchV SR ; TVV SR ). The procedures were reviewed by the committee on animal experiments of the canton of Bern, Switzerland, and approved by the cantonal veterinary authority (Amt für Landwirtschaft und Natur LANAT, Veterinärdienst VeD, Bern, Switzerland) under the license numbers BE/15, BE/17, and BE/

Preparation of figures

Figures were prepared using FlowJo version 10 (FlowJo LLC, Ashland, OR), GraphPad Prism versions and for Windows (GraphPad Software, San Diego, CA), R version , and Inkscape (eunic-brussels.eu).

Statistical analysis

Fold change in median fluorescence intensity (MFI; phosphoflow cytometry) was tested for statistical significance using multiple paired t-tests on log-transformed MFI values of PBS-incubated versus stimulated samples. Obtained p-values are shown alongside q values (two-stage step-up, Benjamini, Krieger, and Yekutieli; desired FDR 5%) and adjusted p-values (Holm-&#x;ídák method).

Differences in glycolytic and mitochondrial ATP production rates (Seahorse assays) were tested for statistical significance using multiple paired t-tests on means of duplicates/triplicates across four assays. Adjusted p-values were determined by the Holm-&#x;ídák method. Calculations were done using GraphPad Prism version for Windows.

Results

Phenotype and transcriptome of nonclassical and intermediate monocytes differ markedly from classical monocytes

Bovine monocyte subsets were analyzed by flow cytometry for expression of various surface molecules (Figures 1A, B). While some molecules showed clear subset-dependent expression patterns (e.g. CD, CD11a, CD11b, CD), other molecules varied strongly with the animals analyzed (e.g. CD40, CD80, MHC-II). A gradual increase from cM over intM to ncM could be observed for expression of CD11a, whereas a gradual decrease was observed for CD11b. In addition to CD11a, ncM expressed the highest levels of CD5, CD8α, and CD Intermediate monocytes expressed the highest levels of CD86 and, for 2 out of 3 animals, also expressed the highest levels of MHC-II. Highest expression on intM combined with lowest expression on ncM was seen for CD and CD11c. Expression of CDa, CD45 and CD43 was higher on ncM and intM compared to cM. Furthermore, cM expressed the highest levels of CD62L, being almost absent from ncM. Supplementary File 1 illustrates exemplary flow cytometry data as well as marker expression on CD14^highCD16^dim monocytes, which were not included in any sorting gates for bulk RNA-seq in the present study. These cells showed expression levels in-between CD14^highCD16^- (cM) and CD14^highCD16^high (intM) monocytes for most markers analyzed, except for CD, CD11b and CD11c, which were expressed on CD14^highCD16^dim cells at levels at least as high or even higher than on cM of the same animal.

Figure 1 Phenotype and transcriptional clustering of bovine monocyte subsets. (A, B) Flow cytometric analysis. Freshly isolated PBMC were stained for flow cytometry. Monocyte subsets were gated based on expression of CD14 and CD16 within CDa^high cells after gating on large cells (FSC^high), single cells (within diagonal in FSC-A vs. FSC-H and SSC-A vs. SSC-H), and living cells (Near-IR^low). Classical monocytes (cM) were gated as CD14^highCD16^-, intermediate monocytes (intM) as CD14^highCD16^high, and nonclassical monocytes (ncM) as CD14^-/dimCD16^high. (B) Graphs show the delta median fluorescence intensity (MFI) of surface expression for selected molecules. Delta MFI was calculated as the difference in MFI between stained samples and FMO controls. Lines illustrate the delta MFI across monocyte subsets. Stainings were performed on seven different animals, resulting in 3 animals analyzed per marker. Within single graphs, data of three different animals is illustrated by solid, dashed, and dotted lines. (C) First two axes of a principal component analysis (PCA) including the most variable genes. Illumina sequencing was performed on RNA isolated from sorted monocyte subsets of three animals. Each dot represents one sample, with the color coding for different cell subsets and the shape coding for the three different animals.

Taken together, our phenotypic analyses have confirmed previously described expression patterns of MHC-II, CDa, CD62L, CD11a and CD11b on bovine monocyte subsets (7) and provide new information on the expression of CD45, CD43, CD5, CD40, CD80, CD86, CD11c, CD, and CD Moreover, in contrast to a previous study, where CD8α expression was concluded to be absent from all monocytes (34), we found indications of weak CD8α expression on ncM and intM. A recent paper, citing the pre-print version of the present manuscript (35), summarizes phenotypic characteristics of bovine monocyte subsets and gating strategies (36).

Principal component analysis of previously published bulk RNA-seq data (9) revealed that differences between monocyte subsets explained the highest proportion of total variance (73%, PC1) in the transcriptomic dataset, followed by 9% of variance (PC2) explained by differences between animals (Figure 1C). As expected from previous analyses (9), ncM and intM clearly clustered apart from cM, with intM clustering much closer to ncM than to cM. Notably, when looking at PC1, intM of animal #2 clustered closer to ncM samples than to intM samples of the other two animals.

Identity of monocyte subsets was confirmed by subset-specific transcription of the key genes NR4A1 (ncM, intM), CX3CR1 (ncM, intM) and CCR2 (cM) (Supplementary File 2A). Moreover, the transcription of surface molecules previously analyzed by flow cytometry in different animals followed the same patterns. Except for CD, for which mRNA content in intM was lower than expected, and CD11b (ITGAM) and CD43, for which mRNA content in ncM was higher than in intM (Supplementary File 2B). Diverging patterns were also observed for CD80, with considerable individual variation both on mRNA and protein level.

Pairwise comparisons of monocyte subsets revealed a variety of differentially expressed genes, involved in various immune functions. The genes addressed in the following chapters have been selected based on pairwise comparisons (DESeq2; adjusted p-value < ) and literature research, and raise no claim for completeness. The output from pairwise comparisons, including normalized counts for all genes and subsets, is provided in Supplementary File 3.

Pro-inflammatory gene expression prevails in classical monocytes

Monocyte subsets clearly differed in the transcription of inflammatory cytokines and cytokine receptors (Figure 2A). Classical monocytes were strongly enriched in transcripts for IL-1 (IL1A, IL1B) and for the IL-1 receptor (IL1R1, IL1RAP). Moreover, cM predominantly expressed IL6R, IL15RA, IL17RA, IL17RC, IL17RD and IL27RA. Trans-signaling with soluble IL-6 receptor is reported to mediate pro-inflammatory functions of IL-6 (37) and also in vitro stimulation with IL is reported to increase inflammasome activation in monocytes (38). Furthermore, expression of TNFSF13 (APRIL), reported to induce IL-8 production in the human macrophage-like cell line THP-1 (39) and TNFRSF1A (TNFR1), mediating pro-inflammatory signaling of TNF-α, was highest in cM. Notably, TNF was found to be primarily expressed in ncM and intM. Tumor necrosis factor alpha (TNF/TNF-α) is regarded as a master regulator of inflammation with various effects on different cell types. TNF-α is produced either in soluble form which signals mainly via TNFR1 (TNFRSF1A), or as transmembrane protein, being the main ligand for TNFR2 (TNFRSF1B) (40) and implicated in reverse signaling (41), the functional consequences of which are incompletely understood (41, 42). Moreover, receptors for IL and IL (IL12RB1, IL12RB2 and IL20RA, IL20RB), were mainly expressed in ncM, though reads for IL20RB were very low (mean=20). Although little is known about ILreceptor- and ILreceptor signaling in monocytes, in vitro studies suggest overall pro-inflammatory effects (43, 44).

Figure 2 Pro-inflammatory gene expression. Illumina sequencing was performed on RNA isolated from sorted monocyte subsets (cM, intM, ncM) of three animals. Heatmaps show row z-scores calculated from log2-transformed normalized counts of selected genes coding for pro-inflammatory cytokines and receptors (A), pro-inflammatory chemokines and receptors (B), proteins associated with the complement system (C), and other pro-inflammatory mediators (D). Mean kilo reads for each subset and gene are given to the right of each heatmap. Genes were selected based on pairwise comparisons with DESeq2 (adjusted p-value < ) and literature research.

Also a number of pro-inflammatory chemokines and chemokine receptors were found to be differentially expressed among monocyte subsets (Figure 2B). Looking at chemokines overexpressed in cM, the most pronounced differences between cM and ncM were found in the expression of CXCL2 and CXCL8. Moreover, transcripts for CCL8, CXCL3, CXCL4 and CXCL6 could be detected in cM and in intM of two animals, though all with low number of reads (max. mean reads). Chemokine receptors associated with inflammation were all predominantly expressed by cM. Expression of CCR1 in cM was upregulated 2-fold over intM and fold over ncM. Expression of CCR2 and CXCR4 was clearly upregulated in cM over intM (fold and 5-fold, respectively), and almost absent from ncM. Exclusive expression in cM was observed for CXCR1 and CXCR2. Nonclassical monocytes and intM clearly showed the highest expression of CCL3 and were also enriched in transcripts for CCL4 and CCL5, though at a lower level.

Gene expression also supports a prominent role of cM in complement-mediated inflammatory processes (Figure 2C). We found that transcription of FCN1, a recognition receptor for the lectin complement pathway, was highly increased in cM and almost absent in ncM, as was the transcription of C1R, a subunit of the complement C1. Notably, transcripts for complement component C3, which is central for activation of both the classical and the alternative complement pathway, were enriched in cM and ncM, and showed the lowest levels in intM, whereas complement factor I (CFI, C3b-Inactivator), an important negative regulator of both complement pathways, was exclusively transcribed in cM and intM. Moreover, transcription of C7 as well as of C8G, both involved in formation of the membrane attack complex, was highest in cM. Nevertheless, certain important genes of the complement system were overexpressed in both ncM and intM. Among those genes were C1QA, C1QB, and C1QC, all of which were barely expressed in cM. The molecule C1q is one of the main sensors of PAMPs and DAMPs, but also antibody complexes, in the classical complement pathway and has been associated with tolerogenic functions (45). Notably, C1q also binds to the surface of dead cells, thereby promoting their phagocytosis. Complement factor 2 (C2) and complement factor D (CFD), involved in the classical and alternative complement pathway, respectively, showed the highest transcription in intM. The receptor for complement factor C3a (C3AR1) was exclusively expressed in intM and ncM, whereas transcription of the receptor for complement factor C3d (CR2/CD21) was markedly increased in ncM and to a lesser extent in intM, when compared to cM. Both known receptor genes for complement factor C5a, C5AR1 (CD88) and C5AR2, showed increased expression in ncM and intM, with slightly higher expression in intM. While C5aR1 is regarded as a mediator of pro-inflammatory signaling, C5aR2 has recently received attention as a multifaceted modulator of C5a signaling, described to dampen inflammasome activation and to alter TLR signaling (46). Furthermore, we found a higher transcription of CD55 in ncM/intM compared to cM, encoding a complement inhibitory protein reported to suppress T-cell responses (47). A fold increased transcription in ncM as compared to cM was found for CD59, encoding a receptor for C8 and C9. Apart from its function as an inhibitor of the membrane attack complex, CD59 expression on antigen-presenting cells was reported to deliver suppressive signals to murine CDexpressing CD4 T cells via a complement-independent ligand (48, 49). Nonclassical monocytes were also clearly enriched in transcripts for the gene ENSBTAG A protein query revealed the highest similarity with human (%) and murine (%) C4BPA, a molecule implicated in the inhibition of classical complement activation that has been reported to induce an anti-inflammatory state in monocyte-derived dendritic cells (50). However, no transcripts were found for the gene annotated as bovine C4BPA.

Also transcription of other genes associated with pro-inflammatory functions was clearly dominant in cM (Figure 2D). This includes genes coding for sensory components of inflammasomes (NOD1, NLRC4, NLRP1, NLRP3), pyroptosis-mediating gasdermin E (GSDME), enzymes involved in the biosynthesis of leukotriens (ALOX5AP, LTA4H) and prostaglandins (PTGES2), as well as the kinase for generation of sphingosinephosphate (SPHK1), the pro-inflammatory receptor TREM1, the pro-inflammatory amidase NAAA (51, 52), and S proteins promoting inflammation (SA7, SA8, SA9, SA12) (53). Furthermore, transcripts for vimentin (VIM), reported to be a key positive regulator of the NLRP inflammasome (54), were clearly enriched in cM. While none of the monocyte subsets contained transcripts for nitric-oxide synthases at steady state (data not shown), we could recently show that bovine cM massively increase transcription of NOS2 upon in vitro stimulation with TLR ligands (24). Nonclassical monocytes expressed the highest levels of CASP4 and GBP5, the latter being described as an activator of inflammasome assembly (55). Caspase 4 (CASP4), being part of the non-canonical inflammasome, is described to promote pro-inflammatory cytokine production, but recently has also been implicated in autophagy (56) &#x; a process that may negatively regulate inflammasome signaling. Taken together, the differential expression of pro-inflammatory genes suggests fundamentally different functions of bovine monocyte subsets.

Gene expression and TLR responsiveness indicate complementary functions of cM and ncM in antibacterial and antiviral immunity

Looking at the gene expression of pattern recognition receptors, we found that TLR2 was expressed higher in cM and intM compared to ncM, and that TLR4 and TLR5 transcripts were clearly enriched in cM (Figure 3A). Toll-like receptor 6 (TLR6) showed a trend towards higher expression in intM and ncM. Expression of TLR3 differed markedly between animals, but was in tendency highest in intM. Furthermore, cM expressed the highest levels of TLR7 and STING1, the latter encoding a cytoplasmic receptor for DNA of both viral and bacterial origin. For two out of three animals, TLR9 expression was also highest in cM. Transcript levels for RIG-1 (DDX58) and MDA-5 (IFIH1), however, were higher in ncM and intM.

Figure 3 Antimicrobial gene expression and TLR responsiveness. Illumina sequencing was performed on RNA isolated from sorted monocyte subsets (cM, intM, ncM) of three animals. (A) Gene expression for pattern-recognition receptors. Graphs show the number of reads across monocyte subsets for selected genes with individual animals indicated by solid (#1), dashed (#2) and dotted (#3) lines. (B, C) Heatmaps show row z-scores calculated from log2-transformed normalized counts for genes associated with antibacterial (B) and antiviral (C) responses. Mean kilo reads for each subset and gene are given to the right of each heatmap. Genes were selected based on pairwise comparisons with DESeq2 (adjusted p-value < ) and literature research. (D) Responsiveness of bovine monocyte subsets to TLR ligands. Defrosted bovine PBMC were depleted of CD3⁺cells, stained for CDa, CD14 and CD16, and stimulated with Pam2CSK4, Poly(I:C), LPS, Gardiquimod, or Resiquimod for 15 min, before being fixed/permeabilized and stained with a fluorochrome-conjugated monoclonal antibody against phosphorylated p38 MAPK. Incubation with PBS served as control. Graphs show the fold change in median fluorescence intensity (MFI of stimulated sample divided by MFI of PBS control) of phospho-p38 MAPK staining for cM (CD14^highCD16^-), intM (CD14^highCD16^high) and ncM (CD14^-/dimCD16^high). For each stimulation, data of four different animals (color-coded dots) is shown. Boxes indicate minimum, maximum and median values. Paired t-tests were performed on log-transformed MFI values of stimulated samples vs. PBS-incubated samples. Statistical analyses and exemplary flow cytometry plots are given in Supplementary File 4.

Bovine cM were clearly enriched in transcripts involved in antibacterial responses (Figure 3B). These transcripts encode an accessory protein for TLR4 (MD2), other LPS-binding proteins [AOAH (57), CRISPLD2 (58)], beta-defensins (DEFB1, DEFB3, DEFB6, DEFB7, DEFB10, DEFB) (59), and other proteins commonly known or described to be involved in antibacterial responses (RNASE6 (60), BPI (61, 62), HP (63), CHI3L1 (64), LYZ). Notably, high levels of MD2 were also expressed by intM of two animals. Three genes associated with antibacterial responses were found to be upregulated in ncM and intM &#x; ACOD1, CD and LY86 (MD1), the latter two genes coding for LPS-binding proteins and members of the TLR family that form a complex to regulate TLR4 signaling (65). ACOD1 (IRG1) mediates the production of itaconate, which is &#x; apart from its anti-inflammatory functions &#x; also known for its antibacterial properties (66).

Looking at genes associated with antiviral responses, we found one gene upregulated in cM (PTPN22) and the vast majority of genes upregulated in intM and ncM (Figure 3C). PTPN22, overexpressed in cM, has been reported to potentiate TLR-induced type-I interferon production (67), and to regulate inflammasome activation (68). Both ncM and intM clearly expressed the highest levels of IRF4 and IFNAR2, whereas IFNAR1 was expressed to similar levels in all monocyte subsets (data not shown). Also IFNGR1 and IFNGR2, coding for the IFN-γ receptor, showed increased transcription in ncM and intM. Accordingly, the transcription of interferon-induced antiviral genes (ISG15, RSAD2, IFIT1, IFIT2, IFI47) was higher in ncM and intM, as compared to cM. Notably, the ubiquitin-like protein ISG15 exerts its antiviral function intracellularly by ISGylation of viral proteins and also extracellularly by acting in a cytokine-like manner to promote IFN-γ production of NK cells and T cells (69). Viperin (RSAD2), a multifunctional antiviral factor (70), has recently gained attention, as it was shown to act as a synthase for antiviral ribonucleotides (71).

Moreover, ncM and intM were enriched in transcripts for several interferon-induced guanylate-binding proteins (GBP1, GBP4, GBP5, GBP6, ENSBTAG). Alongside the GBP4 gene displayed in the heatmap (ENSBTAG), also three other genes annotated as GBP4 (ENSBTAG, ENSBTAG) or GBP4-like (ENSBTAG) were upregulated in ncM and intM (data not shown). Only recently, GBP1 has been allocated an important role in apoptosis and pyroptosis of human macrophages (72). GBP4 has been reported to negatively regulate virus-induced type I IFN responses by targeting interferon regulatory factor 7 (73). Along this line, BST2, exclusively expressed in ncM and intM, was reported to inhibit type I interferon and cytokine production in TLR7/9-stimulated pDC (74). Furthermore, SEC14L1, more than 2-fold enriched in ncM and intM, was reported to negatively regulate RIG-I-mediated signaling (75).

Responsiveness of monocyte subsets to TLR stimulation, as determined by phosphoflow cytometry for p38 MAPK (Figure 3D), corroborates a specialization of cM and ncM for antibacterial and antiviral responses, respectively. While the TLR4 ligand LPS induced responses primarily in cM and intM, stimulation with the TLR7/8 ligands Gardiquimod and Resiquimod induced the strongest responses in ncM. This may point towards high TLR8 expression in ncM (TLR8 not annotated), as transcript levels for TLR7 were relatively low in ncM. Also responses to Pam2CSK4, a synthetic diacylated lipopeptide and ligand for TLR2/6, were highest in ncM. At last, similar low-level responses to Poly(I:C) were detectable in all subsets, with one animal standing out by a considerably higher response (statistical analyses and exemplary flow cytometry plots are given in Supplementary File 4). Taken together, these results indicate that bovine monocyte subsets are specialized in responding to different pathogen-associated molecular patterns, suggesting a role of ncM in immunity against viruses (TLR7/8) as well as gram-positive bacteria and mycoplasma (TLR2/6) (76) and for cM and intM against gram-negative bacteria (TLR4).

Nonclassical monocytes have a gene expression signature promoting resolution of inflammation and tissue repair

All three monocyte subsets expressed anti-inflammatory genes, but ncM and intM were clearly dominant in this regard (Figure 4A).With the expression of anti-inflammatory genes, cM seemed to mainly regulate their own pro-inflammatory functions. Among those genes are regulators of inflammasome activation (MEFV (77), NLRP12 (78)), an enzymatic inactivator of leukotriene B4 (PTGR1) and sphingosinephosphate (SGPL1) (79), a cytokine-scavenging protein (A2M) (80), as well as a negative regulator of cytokine signaling (SOCS3) and nitric oxide production (SPSB2). Moreover, cM were strongly enriched in transcripts for the IL-1 receptor antagonist (IL1RN) and exclusively expressed the decoy receptor for IL-1 (IL1R2). Also IL4R and IL13RA1 were expressed to higher levels in cM, suggesting that cM are especially receptive for IL, which was shown to inhibit the production of pro-inflammatory cytokines in macrophages (81). Furthermore, classical signaling through the IL-6 receptor (IL6R), 3-fold enriched in cM over ncM (Figure 2A), is reported to mediate anti-inflammatory effects of IL-6, as opposed to signaling through soluble IL-6 receptor (37). Notably, while cM expressed the highest levels of IL10, transcripts for the IL receptor (IL10RA, IL10RB) were clearly enriched in ncM and intM. Additionally, both ncM and intM contained IL17REL transcripts, coding for a soluble receptor and potential negative regulator of IL signaling (82). Moreover, ncM and intM were enriched for IL21R transcripts, with IL signaling described to enhance SOCS gene expression and to limit cytokine production in human monocyte-derived cells (83). Also transcripts for TRAIL (TNFSF10), TNFR2 (TNFRSF1B) and DR6 (TNFRSF21) were strongly overexpressed in ncM and intM. Apoptosis-inducing TRAIL expression on monocytes is suggested to be critical for regulation of inflammation (84), and high expression of TNF receptor 2 (TNFRSF1B) in intM and ncM may favor suppressive signaling of transmembrane TNF-α, as reported for murine myeloid-derived suppressor cells (85). Death receptor 6 (TNFRSF21) has been reported to have inhibitory effects on monocyte differentiation when cleaved from the surface of tumor cells by matrix metalloproteinase 14 (MMP14) (86), the latter being expressed about fold higher in ncM and intM (data not shown).

Figure 4 Gene expression associated with anti-inflammatory responses and tissue repair. Illumina sequencing was performed on RNA isolated from sorted monocyte subsets (cM, intM, ncM) of three animals. Heatmaps show row z-scores calculated from log2-transformed normalized counts of selected genes associated with anti-inflammatory responses (A), and tissue repair (B). Mean kilo reads for each subset and gene are given to the right of each heatmap. Genes were selected based on pairwise comparisons with DESeq2 (adjusted p-value < ) and literature research.

Other anti-inflammatory genes over-expressed by ncM and intM are reported to be mostly involved in negative regulation of NF-κB signaling (NUMBL, TNIP3, TRIB3 (87), ZCCHC11, DYNC2I2 (88), ZFAND6 (89), MTURN (90), HIVEP3 (91), DUSP5, LRRC25, COMMD9 (92), ADCY7 (93)), but also include surface receptors involved in regulation of inflammation, such as GPR18 (receptor for resolvin D2) (94), SLAMF7 (95), SUCNR1 (96), and CDA (97).

Notably, expression of ACOD1 was over 6-fold higher in ncM compared to cM. The metabolite itaconate, generated by IRG1 (ACOD1), is well-known for its anti-inflammatory effects (66, 98). Furthermore, BCL2 and BCL6 were transcribed to higher levels in ncM and intM. BCL-2 was shown to negatively regulate caspase-1 activation (99) and BCL-6 was recently reported to exert anti-inflammatory effects by suppressing IL6 transcription in murine macrophages (). Like in human CD16⁺ monocytes (), HMOX1 was significantly increased in ncM and intM. Heme oxygenase-1 (HMOX1) was shown to be induced by IL and to mediate the anti-inflammatory effect of IL in murine macrophages, presumably via NF-κB suppression by the heme degradation product carbon monoxide (). Also in human monocytes, HMOX1 was reported to inhibit LPS-induced TNF-α and IL1-β production (). Additionally, RAB7B, described to promote degradation of TLR4 () and TLR9 (), was more than 2-fold higher expressed in ncM and intM. As was LYN, also expressed in human ncM and intM (, ), and reported to negatively regulate TLR-induced cytokine responses (). Notably, LYN has recently been proposed as a negative regulator of murine ncM development (). Overall, these results indicate anti-inflammatory functions for ncM and intM.

In line with their suggested anti-inflammatory and pro-resolving functions, a number of genes associated with different stages of tissue repair were upregulated in ncM and intM (Figure 4B). While cM were enriched in F13A1, mediating hemostasis (), intM and ncM expressed SERPINE1, described to regulate clot resolution (). Both ncM and intM contained the highest transcript levels of genes associated with efferocytosis [AIF1 (), CDLB (), MAFB (), MERTK (), PECAM1 (), VAV3 ()]. Notably, MERTK has also been described as a negative regulator of human T-cell activation ().

Furthermore, ncM expressed higher levels of TGFB1 and of genes coding for TGF receptors (TGFBR2, ACVRL1) or being directly or indirectly involved in TGF-β pathways (ACE, GUCY1B3, ITGB2, SMAD3). Transforming growth factor beta (TGF-β) is known as a pro-fibrotic cytokine involved in wound healing (). Notably, cM expressed higher levels of the TGF receptor genes TGFBR1 and TGFBR3, and of TGFBI and TGIF1. Genes associated with extracellular matrix components dominantly transcribed by ncM included HSPG2 (basement membrane-specific heparin sulfate) and NDST1 (biosynthesis of heparin sulfate), whereas cM contained the highest transcript levels for collagens COL14A1, COL7A1, as well as for EMILIN1 (associates with elastic fibers) and VCAN (versican). Furthermore, fibroblast growth factor FGF9 and platelet-derived growth factor PDGFD were clearly enriched in ncM. The high expression of CMLKR1 in intM and ncM suggests that these subsets are attracted to inflamed tissues via chemerin and contribute to resolution of inflammation via binding to resolvin E1 (), which was shown to increase IL production and phagocytosis of apoptotic neutrophils in macrophages (, ).

In addition, several metalloproteinases and their regulators were either upregulated in ncM (ADAM9, ADAMTSL5, ADAMTS12, MMP14, MMP19, TIMP2, TSPAN14, ECM1) or cM (ADAM19, ADAM8, ADAMTS2, MMP25) (data not shown). Several metalloproteinases are described to be involved in wound healing (), among which MMP14 and ADAM9 are suggested to regulate epithelial cell proliferation (, ). Finally, several genes associated with angiogenesis (ADGRE5/CD97 (), ADTRP, ANXA3 (), CEACAM1 (), CMKLR1 (), ECM1 (), FMOD (), SEMA7A (), VASH1) were predominantly expressed by intM and ncM. Notably, SEMA4A (), also reported to be involved in angiogenesis, showed the highest transcription in cM. Altogether, these data suggest anti-inflammatory pro-resolving functions of ncM and intM, as well as a prominent role of these subsets in wound healing and tissue regeneration.

Gene expression indicates differential capabilities for antigen presentation, co-stimulation, and modulation of T-cell responses

Antigen presentation capabilities are reported for monocytes across species (1). As expected from phenotypic analyses, intM stood out by their high gene expression for MHC-II (BOLA-DQA5, BOLA-DQB, BOLA-DRA), and the co-stimulatory molecules CD40 and CD86 (Figure 5A). Interaction of CD40 with CD40L on T cells has been reported to stimulate Th17 responses (). Notably, genes for MHC class I molecules (BOLA, BoLA), were transcribed to higher levels in intM and ncM. Furthermore, mRNA from genes associated with the presentation of lipid antigens to T cells (ENSBTAG annotated as CD1a molecule-like, CD1E), was enriched in intM.

Figure 5 Expression of genes involved in the shaping of T-cell responses. Illumina sequencing was performed on RNA isolated from sorted monocyte subsets (cM, intM, ncM) of three animals (#). (A) Gene expression promoting antigen presentation and co-stimulation. Graphs show the number of reads across monocyte subsets for selected genes with individual animals indicated by solid (#1), dashed (#2) and dotted (#3) lines. (B) Gene expression involved in T-cell modulation. Heatmap shows row z-scores calculated from log2-transformed normalized counts of selected genes. Mean kilo reads for each subset and gene are given to the right of the heatmap. Genes were selected based on pairwise comparisons with DESeq2 (adjusted p-value < ) and literature research.

Genes encoding T-cell signaling cytokines predominantly expressed by cM included IL12B (Figure 5B), as well as TNFSF8 (CD30L) and TNFSF14 (LIGHT) (, ). In fact, transcription of IL12B mRNA was found to be absent in ncM, and over fold increased in cM over intM. Two further genes upregulated in cM and reported to be involved in regulation of T-cell activation include CLECL1 and TARM1. CLECL1, fold increased over ncM, has been reported to act as a T-cell costimulatory molecule, skewing the CD4 T-cell response towards Th2 by increasing IL-4 production and proliferation (). TARM1, about 6-fold enriched in cM, has been reported to suppress CD4-T-cell activation and proliferation in vitro ().

T-cell signaling cytokines predominantly expressed by intM and ncM included IL7 and IL15, reported to function in lymphoid homeostasis (), and IL27, encoding a multifaceted cytokine described to both promote and suppress T-cell responses (). Also, EBI3 (ILβ), an essential component of the cytokines IL and IL, was exclusively expressed in intM and ncM (approx. fold increased). Moreover, TNFSF9, a ligand for CD on T cells shown to be important for the generation of antiviral CD8-T-cell responses (), was 8-fold enriched in ncM over cM.

Furthermore, IL11RA (alpha subunit of the IL receptor) was expressed at higher levels in ncM and intM. In line with almost absent IL12B transcription in ncM and intM, IL signaling has been reported to inhibit IL production in macrophages (), which supports polarization towards Th2 responses (). As shown in Figures 2B, 5B, chemokines relevant for T-cell responses were mainly expressed by ncM and intM (CCL3, CCL4, CCL5 and CCL16, CXCL10, CXCL16).

In fact, the vast majority of genes (22 out of 27) associated with modulation of T-cell responses was overexpressed in ncM and intM (Figure 5B). Intermediate monocytes expressed the highest levels of adenosine deaminase (ADA), which is reported to act as a modulator of T-cell differentiation, increasing the generation of effector, memory, and regulatory T cells (). Both ncM and intM were enriched in transcripts for SLAMF6, reported to boost IFN-γ production and cytolytic anti-tumor activity of human CD8 T cells in vitro (). Furthermore, expression of ALCAM, encoding a ligand for CD6 on T cells, important for stabilizing the immunological synapse between APC and T cells () and reported to mediate extravasation of monocytes (), was more than 3-fold higher expressed in ncM than in cM. Notably ADGRE1 (F4/80), reported to be essential for the generation of Tregs and peripheral tolerance when expressed on antigen-presenting cells (), was fold enriched in ncM over cM. Moreover, ADGRE5, coding for CD97 and described to induce regulatory T cells and IL production upon engagement of CD55 on CD4 T cells (, ) was 3-fold higher transcribed in ncM and intM. Consistent with the idea that ncM promote the generation of Tregs, they showed the highest transcription of BTN2A2, a butyrophilin reported to inhibit activation and induce Foxp3 expression in murine T cells (, ).

Many T-cell modulating genes overexpressed in ncM and intM were found to be genes involved in negative regulation of T-cell activation (BTN2A2, CD52, CD83, CD/PDL1, PDL2, IDO1, PECAM1, and VSIG4). Soluble CD52 was reported to suppress T-cell activation via binding to Siglec () and soluble CD83 was shown to regulate T-cell activation by binding to the TLR4/MD-2 complex on human CD14⁺ monocytes and inducing expression of anti-inflammatory mediators such as IDO and IL (). Transcription for PDL-1 (CD), a well-known inhibitor of T-cell activation (), was increased 6-fold in ncM over cM. Notably, PDL-1 has recently been employed as a marker of ncM for in-vivo tracking of this monocyte subset in mice (). Similarly, the gene for PDL-2, a second ligand for PD-1 on T cells with T-cell inhibitory function (), was 9-fold higher expressed in ncM, though at lower levels than the gene for PDL-1 (CD). Strikingly, IDO1 expression was found to be significantly increased in intM (fold) and ncM (fold) when compared to cM, where expression was almost absent (mean of 20 reads). IDO1 was described to inhibit T-cell activation by degrading tryptophan, and to promote tolerance of DC and the expansion of Tregs (). Also transcription of PECAM1 (CD31), described as a key co-inhibitory receptor promoting tolerogenic functions in both DCs and T cells through homophilic interactions (), was greatly upregulated in ncM and intM. A recent in vitro study also suggested that high CD31 expression on DCs reduces priming of CD4 T cells by impairing stable cell-cell contacts (). Furthermore, VSIG4, coding for a B7-family related protein specifically expressed on resting macrophages was primarily expressed in ncM. Notably, VSIG4 has been described as a strong negative regulator of T-cell activation, maintaining T-cell unresponsiveness in healthy tissues (). Taken together, these data clearly indicate that monocyte subsets are actively involved in the shaping of T-cell responses, with ncM and intM being especially well equipped for T-cell suppression, either directly or via the induction of regulatory T cells.

Expression of metabolic genes differs markedly between classical and nonclassical monocytes

Given that metabolism and immune function are tightly linked (&#x;), differences in metabolic pathways can give indications for subset-specific functions. In fact, monocyte subsets showed prominent differences in the expression of genes relating to metabolism. The vast majority of these differentially expressed genes was associated with glycolysis and showed the highest expression in cM (Figure 6A). Apart from glycolytic genes, also genes involved in oxidative phosphorylation showed increased transcription in cM as compared to ncM (Figure 6B). Among those genes was ATP5ME (subunit of the mitochondrial ATP synthase), a mitochondrial inner membrane protein (MPV17) supporting oxidative phosphorylation (), the genes coding for the components of complex II of the mitochondrial electron transport chain succinate dehydrogenase (SDHA, SDHB, SDHC, SDHD), and also most of the numerous genes coding for subunits of complex I (data not shown). In line with increased glycolysis in cM, transcripts for the glucose transporters SLC2A1 and SLC2A3 were about 3-fold enriched in cM (data not shown). As reported in our previous publication (9), also SLC genes involved in transport of succinate (SLC13A3), citrate (SLC13A5) and lactate/pyruvate (SLC16A1) were mainly expressed in cM. Notably, the metabolites succinate and citrate are both described as critical pro-inflammatory mediators linking metabolism to immune functions (). Moreover, two genes associated with fructose metabolism (KHK, SORD) were predominantly expressed in cM. Fructose-induced metabolic changes were recently described to enhance inflammatory responses of dendritic cells (). In addition, genes involved in the oxidative (G6PD, PGD) and non-oxidative (TALDO1) pentose-phosphate pathway (PPP) showed the highest transcription in cM. The PPP, providing redox-equivalents and nucleotide precursors, was shown to be essential for pro-inflammatory functions of human macrophages (, ). Furthermore, PANK1, coding for a key enzyme in CoA synthesis was 5-fold upregulated in cM over ncM.

Figure 6 Metabolic gene expression and Agilent Seahorse Assays. (A, B) Illumina sequencing was performed on RNA isolated from sorted monocyte subsets (cM, intM, ncM) of three animals. Heatmaps show row z-scores calculated from log2-transformed normalized counts of glycolytic genes (A) and of genes associated with other metabolic pathways (B). Mean kilo reads for each subset and gene are given to the right of each heatmap. Genes were selected based on pairwise comparisons with DESeq2 (adjusted p-value < ) and literature research. (C+D) Bovine monocyte subsets were FACS-sorted and metabolic activity was analyzed by Agilent Seahorse XF technology and the XF Real-Time ATP Rate Assay. Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were automatically calculated from measured oxygen and pH decrease (Agilent Wave software). The electron transport chain inhibitors oligomycin and rotenone/antimycin A were injected sequentially to allow calculation of OXPHOS- as well as glycolysis-mediated ATP production rates (mito, glyco) from resulting OCR and ECAR. Monocyte subsets from four animals were analyzed. (C) For one representative animal, OCR and ECAR traces are depicted in addition to ATP production rates which are shown as absolute values (bar graphs) and relative values (pie charts). (D) ATP production rates for three other animals. Data of intM (2 animals) and monocyte-derived macrophages (2 out of 3 animals) is shown in Supplementary File 6.

Notably, all three genes coding for members of the PPAR family (PPARA, PPARD, PPARG) were transcribed at higher levels in cM. These lipid-activated nuclear receptors have evolved as key regulators linking lipid metabolism to inflammation, and in particular expression of PPARG is associated with anti-inflammatory functions (). We recently showed that bovine cM downregulate PPARG transcription dramatically upon Gardiquimod (TLR7/8) stimulation in vitro (24). PPARA was recently proposed as an important mediator of antimicrobial responses to mycobacteria (), inducing expression and translocation of TFEB, a key transcriptional activator of autophagy and lysosomal biogenesis. Notably, while TFEB was equally expressed in all monocyte subsets (data not shown), transcription of TFEC, a less well described member of the TFE family, was approximately 9- and 6-fold increased in cM and intM respectively compared to ncM.

Altogether, these data suggest that cM are metabolically more active than ncM and intM, with a significantly enhanced expression of genes involved in glycolysis, supporting pro-inflammatory functions.

Mitochondrial respiration prevails in nonclassical monocytes

To investigate differential use of metabolic pathways in monocyte subsets, we assessed the contribution of glycolysis and mitochondrial respiration to their ATP production. Extracellular flux analysis of sorted monocyte subsets demonstrated that cM produced the majority of their ATP through glycolysis, whereas ncM predominantly used oxidative phosphorylation (OXPHOS) for ATP production (Figure 6C, D). Raw data and statistical analyses (ATP production rates) are shown in Supplementary File 5.

Intermediate monocytes appeared to proportionally produce more ATP via glycolysis than ncM (Supplementary File 6), however these results should be interpreted with caution, as bulk RNA-seq of intM revealed pronounced animal-to-animal variability, presumably caused by considerable heterogeneity, or mixed populations within the intM gate.

Monocyte-derived macrophages (cM-M, intM-M, ncM-M) generated by a 6-day in-vitro culture showed considerably higher metabolic activity evident also by the need to reduce the cell number for Seahorse assays by five times. Notably, after this 6-day culture, the proportion of ATP produced by OXPHOS was increased in all subsets (Supplementary Files 6C, D; right panels). It remains to be elucidated whether this switch to OXPHOS is a hallmark of macrophage differentiation or rather a result of in-vitro culture conditions. Certainly, the observed preferences for different metabolic pathways ex vivo are in line with diverging roles of monocyte subsets in inflammation and beyond.

Gene set enrichment analysis

In addition to the manual gene-by-gene analysis described above, automated analyses with pre-defined gene sets were performed for the bulk RNA-seq datasets. Intermediate monocytes were excluded from the analyses due to the high animal-to-animal variability observed for certain genes. Enrichment analysis with gene ontology gene sets related to biological processes was found to be most informative. In line with the manual analysis, genes overexpressed in cM vs. ncM (DESeq2 output; adjusted p-value < ) were enriched (q value < ) in gene sets related to metabolism as well as to antibacterial and pro-inflammatory responses, whereas genes overexpressed in ncM vs. cM were enriched in gene sets related to adhesion, T-cell regulation, wound healing and antiviral responses.

The complete list of gene sets is provided as Supplementary File 7, including the lists of genes assigned to these gene sets, which should be carefully examined before making further conclusions. Especially for genes overexpressed in ncM, many gene sets were found to be misleading, suggesting for example an involvement in B-cell responses due to genes expressed in, and related to, B cells.

Single-cell transcriptomic data suggests continuous differentiation

Single-cell RNA-seq has provided unprecedented insights into the heterogeneity of myeloid cells (), highlighting that the classification of human and bovine monocyte subsets according to expression of CD14 and CD16 may be an oversimplification. In order to get an unbiased view on monocyte subset composition, we have performed 10x Genomics single-cell RNA-seq of bovine PBMC (summaries of Cell Ranger outputs are given in Supplementary File 8). Within PBMC, monocytes were readily identified by visualizing SIRPA expression in the UMAP projection (Figure 7A). To better resolve potential subclustering, clusters belonging to this group of SIRPA-expressing cells (c4, c10, c14, c18; Supplementary File 9A) were selected and re-clustered independently from the other cells in the PBMC dataset (Figure 7B). From the resulting eight clusters (resolution ), only six clusters (clusters ) were further analyzed, as clusters 6 and 7 appeared to contain doublets with expression of T- and B-cell associated genes alongside SIRPA expression (Supplementary File 9B). Visualization of key genes and analysis of the top differentially expressed genes identified cluster 2 as ncM/intM and cluster 5 as cDC2 (Figures 7C, D). Clusters 0, 3 and 4 appeared to contain cM in different activation states. Notably, cluster 4 stood out by high expression of several defensin genes. The complete list of differentially expressed genes is given in Supplementary File

Figure 7 Single-cell RNA sequencing reveals heterogeneity of bovine monocytes. PBMC from two cows were subjected to 10x Genomics 3&#x; single-cell RNA sequencing. (A, B) Clusters containing SIRPA-expressing putative monocytes were subsetted for independent re-clustering. Clusters 6 and 7 were excluded from further analysis. (C) Feature plots show expression of SIRPA, CD14, FCGR3A (CD16) and FCER1A in the re-clustered dataset. (D) Heatmap shows the top 5 (adjusted p-value) differentially expressed genes, revealing expression signatures resembling classical monocytes (c0, c3, c4), intermediate monocytes (c1, c2), nonclassical monocytes (c2), and cDC2 (c5). Differential expression testing was performed with seurat&#x;s FindAllMarkers function. (E-G) Visualization of selected signature genes enriched in bulk-sequenced cM (E), intM (F), and ncM (G). (H) Trajectory analysis on clusters and visualization of genes differentially expressed along a selected trajectory (indicated in red). The complete list of differentially expressed genes along this trajectory is given in Supplementary File

Visualization of signature genes derived from bulk-sequenced cM supported their cluster annotation, and for some genes (e.g. LYZ, SA8), also revealed a gradual expression increase from clusters 1 and 0 towards clusters 3 and 4 (Figure 7E), while VIM and F13A1 showed the highest transcription in cluster 0 and 3, respectively. In line with their intermediate nature, transcripts enriched in bulk-sequenced intM (e.g. ANXA3, CFD, CD1E) were mostly detected in cells connecting clusters 1 and 2 (Figures 7B, F). It must be noted that bulk sequenced intM of the present study were defined as CD14^highCD16^high. It is expected that cluster 1 contains CD14^highCD16^dim cells, which were not included in the sorting gate used for bulk RNA sequencing. Transcripts enriched in bulk-sequenced ncM were almost exclusively detected in cluster 2 (Figures 7B, G). Notably, expression of C1QA (Figure 7G), as well as of C1QB and C1QC (not shown) appeared to be restricted to a subcluster within cluster 2, suggesting the presence of distinct cell states within ncM.

Trajectory analyses performed with Monocle 3, resulted in net-shaped trajectories spanning clusters 0, 1, 3 and 4, and a straight trajectory from cluster 1 towards cluster 2. Genes for which expression either significantly (q value < ) decreased or increased along this straight trajectory are shown in Figure 7H (selection) and Supplementary File 11 (complete list).

Taken together, the unbiased analysis of single-cell transcripts revealed remarkable heterogeneity of bovine blood-derived cM in healthy animals, presumably reflecting a continuum of activation states, and supports the hypothesis that ncM are generated via differentiation from cM, with intM (CD14^highCD16^high), or a subpopulation thereof, representing a transient intermediate state, potentially specialized in lipid antigen presentation to T cells (Figure 8). Future studies need to address CD14^highCD16^dim monocytes and the heterogeneity of CD14^highCD16^high intermediate monocytes.

Figure 8 Proposed continuous differentiation and proposed functional specialization of bovine monocyte subsets. (A) Assignment of scRNA-seq clusters to CD14/CDdefined monocyte subsets, as sorted for bulk RNA-seq. (B) Proposed differentiation pathways. A subcluster of cM (c0) may give rise to anti-inflammatory intM and ncM with specialized functions in antigen presentation and antiviral responses, respectively, and to pro-inflammatory subclusters of cM with prominent antibacterial functions.

Discussion

With the present study, we extended the phenotypic characterization of bovine monocyte subsets and combined an in-depth analysis of their bulk- and single-cell transcriptomes with metabolic and TLR-stimulation assays to get detailed insights into subset-specific functions. Pairwise comparison of bulk gene expression coupled with extensive literature research revealed substantial transcriptomic differences between bovine monocyte subsets, likely determining their specializations, while single-cell transcriptomics provided an unbiased view on subset composition supporting differentiation of cM towards ncM via relatively transient intM. Genes differentially expressed between bulk-sequenced cM and ncM were also analyzed by GSEA with GO_BP gene sets. While GSEA largely confirmed the manual analysis, results of GSEA need to be interpreted with caution and require manual validation with respect to gene-set composition.

Bovine cM clearly emerged as pro-inflammatory, with overall gene expression supporting antibacterial inflammatory responses. This is in line with data on human and murine cM, and with earlier studies on bovine cM that have shown their superiority in phagocytosing bacteria (34). Both ncM and intM were dominant in the transcription of many genes associated with regulatory functions. The expression of anti-inflammatory genes and numerous genes associated with wound healing (efferocytosis, angiogenesis, fibrosis) clearly indicate that bovine ncM are specialized in the resolution of inflammation and in tissue regeneration, as suggested for ncM based on studies in mouse models (). In line with our steady-state transcriptomic data, bovine ncM were previously shown to almost lack IL-1β production upon inflammasome activation in vitro (34). However, literature on the ability of human ncM to produce IL1-β is conflicting (6, ).

Pro- and anti-inflammatory functions of cM and ncM, respectively, are also supported by their metabolic transcriptome, clearly indicating that cM are metabolically more active and skewed towards pro-inflammatory glycolysis. Differential use of ATP-generating metabolic pathways could also be confirmed by extracellular flux analysis, where cM predominantly performed glycolysis and ncM mainly employed oxidative phosphorylation (OXPHOS). These metabolic differences are in line with reported lower glucose uptake of bovine ncM as compared to cM (), and match gene expression as well as respirometric measurements in human cells (), suggesting similar metabolic programing and functional specialization of monocyte subsets in humans and cattle.

Transcriptomic data also suggest diverging functions of bovine monocyte subsets in the interaction with T cells. Notably, intM showed the highest expression of MHC-II, both on mRNA and protein level. High expression of MHC-II is also reported for human intM (5, 6), and may be linked to their superiority in stimulating human CD4-T-cell proliferation (5). Also considering the high expression of CD86 and the high transcription of CD1E and a CD1a-like gene, bovine intM may be particularly well equipped for co-stimulation, and lipid antigen presentation to T cells. A specialization in antigen presentation is also reported for human intM (15). Nonclassical monocytes and intM were enriched in transcripts for various genes promoting CD8-T-cell responses. This preferential activation of CD8 T cells and the risk associated with uncontrolled cytolytic T-cell responses might explain why ncM and intM also show high expression of genes mediating the inhibition of T cells and the generation of Tregs, the latter being reported for murine ncM (). Furthermore, gene expression promoting activation of CD8 T cells, together with the interferon-associated gene signature and the high responsiveness to Gardiquimod and Resiquimod, indicate a specialization of bovine ncM towards antiviral responses, as suggested for human ncM (14).

In support of T-cell stimulating functions, we have detected monocytic cells transcriptionally resembling ncM, intM and cM in bovine mesenteric lymph nodes (scRNA-seq; manuscript in preparation), however the mechanisms of lymph-node entry remain elusive. In contrast to ncM and intM, which almost lack CD62L expression, bovine CD62L^high cM should be able to enter lymph nodes directly from blood. Also lymph-mediated entry of antigen-presenting monocytes has been reported for mice (), however we could not observe CCR7 upregulation in/on stimulated bovine monocytes (24), making lymph-mediated entry via CCR7 rather unlikely.

As reported for human intM (6), bovine intM expressed the majority of genes at levels in-between cM and ncM, while showing higher transcriptional similarity with ncM. Notably, bovine intM were reported to produce the highest amounts of reactive oxygen species in response to opsonized bacteria and the highest amounts of IL-1β following inflammasome activation (34). This is surprising when looking at the steady-state transcriptome of intM described in the present study. A major limitation of studies on intM across species is their poor phenotypic definition, likely including multiple subsets or different activation states leading to conflicting results (11). In fact, heterogeneity of intM has been described for both humans () and mice (). The lack of a clear cluster assignment in our scRNA-seq data and the observation that animal-to-animal variability was most prominent for intM in the bulk dataset support the idea that also bovine CD14^highCD16^high intM are a mixed population &#x; or at least contain various cell states. Notably, a reported expansion of intM in response to dengue-virus infection was recently revealed to be an upregulation of CD16 on human cM (). Also for bovine cM, an upregulation of CD16 is reported following stimulation with IFN-γ (34). Similarly, we found that sorted bovine cM (CD14⁺CD16^-) were all CD16^high after overnight culture (unpublished observation), making them phenotypically indistinguishable from intM when using the standard gating strategy with CD14 and CD Therefore, also dominant glycolysis and LPS responsiveness of intM, observed in the present study and both reminiscent of cM, should be interpreted in the light of possible gate contamination with cM that have recently upregulated CD16 expression &#x; potentially a phenotypic alteration that precedes profound transcriptomic alterations in intM.

It is widely accepted that murine and human intM and ncM arise from cM in the periphery (19, , ). Our single-cell RNA-seq data enable an unbiased view on monocyte subset composition in bovine blood and support the idea of sequential differentiation. Along this line, intM &#x; as identified by bulk-derived signature genes &#x; appeared to be bridging between cM and ncM in the UMAP projection of our single-cell dataset. Further studies are required to understand the differentiation trajectories and to elucidate whether bovine ncM can differentiate from cM under steady-state and/or inflammatory conditions and if and how they can enter tissues to fulfill their specialized functions during immune responses induced by infections and vaccinations. Given the striking similarities of bovine and human monocyte subsets (15), insights from studies in cattle should also advance our understanding of monocyte biology and associated diseases in humans.

Data availability statement

The original contributions presented in the study are publicly available. This data can be found here: [eunic-brussels.eu].

Ethics statement

The animal study was reviewed and approved by the committee on animal experiments of the canton of Bern, Switzerland, and the cantonal veterinary authority (Amt für Landwirtschaft und Natur LANAT, Veterinärdienst VeD, Bern, Switzerland) under the license numbers BE/15, BE/17, and BE/

Author contributions

ST and AS designed the study. ST performed laboratory work, analyzed and interpreted the data and wrote the manuscript. GTB performed laboratory work and analyzed flow cytometry data. HL and RB bioinformatically processed scRNA-seq data. RR and LvM assisted with Seahorse experiments and data interpretation. AS assisted with data interpretation and data analysis. All authors contributed to the article and approved the submitted version.

Funding

Open access funding was provided by the University Of Bern.

Acknowledgments

We want to thank the team of the Clinic for Ruminants at the Vetsuisse Faculty in Bern and the team of animal caretakers at the IVI in Mittelhäusern for blood sampling. Many thanks also go to Giuseppe Bertoni and Nicolas Ruggli for obtaining the cantonal licenses for blood sampling. We also want to thank Sylvie Python and Gael Auray for their support with cell sorting and RNA extraction and Corinne Hug for isolating PBMC and preparing monoclonal antibodies. Furthermore, we want to thank Stefan Müller and Thomas Schaffer (Flow Cytometry and Cell Sorting Facility, University of Bern) for cell sorting. Finally, we want to thank Pamela Nicholson and Tosso Leeb from the Next Generation Sequencing Platform of the University of Bern, and Irene Keller (Interfaculty Bioinformatics Unit, University of Bern) for their support with RNA-sequencing and bioinformatic analyses. A previous version of the manuscript has been uploaded to the preprint server bioRxiv (35).

Conflict of interest

Author LvM is employed by Bucher Biotec AG.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher&#x;s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: eunic-brussels.eu#supplementary-material.

References

4. Hume DA, Irvine KM, Pridans C. The mononuclear phagocyte system: The relationship between monocytes and macrophages. Trends Immunol () &#x; doi: /eunic-brussels.eu

PubMed Abstract CrossRef Full Text CrossRef Full Text

16 &#;&#;&#;&#;&#;&#; &#; (CET) &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; "Forex Realtime SA"(&#;&#;&#;&#;&#;&#;&#;&#;&#;). &#;&#;&#;&#; &#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#; - USD, &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#; - USD, &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#; - USD. &#;&#;&#;&#; &#;&#;&#; &#;&#;&#;, &#;&#;&#; &#;&#; &#;&#;&#;&#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;: eunic-brussels.eu

&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; !!!

&#;&#;&#; &#;&#; &#;&#;&#; &#;&#;&#;&#;&#;&#;, &#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;. &#;&#;&#;&#;&#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;, &#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;.

&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; ? &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#; &#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;. &#;&#;&#;&#;&#; &#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;. &#;&#;&#;&#;&#;&#;&#;- &#;&#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;, &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#; &#;&#; Forex &#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;. &#;&#;&#;&#;&#;&#;&#;, &#;&#;&#; &#;&#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#;, &#;&#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;, &#;&#;&#; &#;&#;&#;-&#;&#;&#;&#;&#;&#; &#;&#; &#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;. &#;&#;&#; &#;&#;&#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;, &#; &#;&#;&#;&#;&#; &#; &#;&#;&#; &#;&#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#;&#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;. &#;&#; &#;&#;&#; &#;&#; &#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;, &#;&#;&#;&#; &#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#; USD, &#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#; &#;&#;&#;&#;. &#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;. &#;&#; &#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;, &#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;, &#; &#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;, &#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;, &#;&#;&#; &#; &#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;. &#;&#; &#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;.

&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;, &#; &#;&#;&#;&#;&#; &#;&#;&#; &#;&#;&#;? &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;, &#;&#;&#; &#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#; &#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;. &#;&#; &#;&#;&#;&#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;. &#;&#;&#;&#;&#; &#;&#; &#;&#;&#;&#;&#;&#;, &#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;, &#;&#;&#;&#;&#;&#; &#;&#;&#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;. &#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#; ? &#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#;, &#;&#;&#; &#;&#;&#;&#;&#;&#;&#;, &#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#; 30 &#;&#; &#;&#;&#;&#;&#;&#;&#;. &#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#;, &#;&#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#; &#; &#;&#;&#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;. &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#;&#; &#; &#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;, &#; &#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#; &#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#; FX CONTESTS.

&#;&#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;. &#;&#;&#;&#;&#;&#;&#;, &#;&#;&#; &#;&#;&#;&#; &#;&#;&#;&#;&#;&#; &#;&#; &#;&#;&#; &#;&#;&#;&#; &#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;, &#; &#;&#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;.

&#; &#;&#;&#;, &#;&#;&#; &#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#; &#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;&#;&#;&#; MISCELLANIOUS.

&#; &#;&#;&#;&#;&#;&#;&#; FEEDBACK &#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#; &#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#;&#;&#;, &#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#; &#; &#;&#;&#;&#;, &#;&#;&#; &#; &#; &#;&#;&#;&#;&#; &#;&#;&#;&#;&#;&#; &#;&#;&#; &#;&#;&#;&#;&#;&#;&#;&#;&#;&#;.

&#;&#;&#;&#;&#;&#; &#;&#;&#; &#;&#;&#;&#;&#;&#;&#;: [email protected]

Google Scholar

Narasimhan PB, Marcovecchio P, Hamers AAJ, Hedrick CC. Nonclassical monocytes in health and disease. Annu Rev Immunol () &#x; doi: /annurev-immunol

PubMed Abstract

nest...

2730 2731 2732 2733 2734

как торговать на форексе стратегии индикаторы демарка для metastock календарь новостей на форекс иностранные банки форекс как выставлять стоп лосс на форексе как зарегистрироваться на рынке форекс индикаторы цифровые световые типа ицс 5 в авиации масд на форексе матожидание на форексе все о рынках форекс lang ru индикаторы на omron m2 eco индикаторы в метрологии медицинских учреждениях как заработать деньги на форексе с помощью мета трейдера

Дц Юмис Форекс

Toll-like receptor signaling in thymic epithelium controls monocyte-derived dendritic cell recruitment and Treg generation

Introduction

Results

mTECs express a set of TLRs and signaling adaptors

MyDdependent chemokine expression in mTECshigh

TLR9/MyD88 signaling in mTECs targets Sirpα+ cDC2

TLR9/MyD88 signaling in mTECs recruits CD14+moDCs

TLR9/MyD88 signaling in mTECs affects Treg development

Discussion

Methods

Mice

Tissue preparation and cell isolation

Flow cytometry analysis and cell sorting

Imaging flow cytometry

In vitro TLRs and cytokines stimulation assays

In vivo TLR stimulation

Immunofluorescent analysis of thymic cryosections

Gene expression analysis by qRT-PCR

Bone marrow chimera generation

RNA-sequencing and analysis

Single-cell RNA sequencing

Single-cell RNA sequencing analysis

In vitro antigen presenting assay

Induction of T cell transfer colitis and histological analysis

In vitro Tregs suppression assay

In vivo model of autoimmune diabetes

Statistical analysis

Reporting summary

Data availability

Change history

References

Monocyte biology conserved across species: Functional insights from cattle

Introduction

Materials and methods

Isolation of bovine PBMC

Phenotyping of monocyte subsets by flow cytometry

Fluorescence-activated cell sorting and bulk RNA sequencing

Phosphoflow cytometry

Extracellular flux analysis

Single-cell RNA sequencing (10x Genomics)

Ethics Statement

Preparation of figures

Statistical analysis

Results

Phenotype and transcriptome of nonclassical and intermediate monocytes differ markedly from classical monocytes

Pro-inflammatory gene expression prevails in classical monocytes

Gene expression and TLR responsiveness indicate complementary functions of cM and ncM in antibacterial and antiviral immunity

Nonclassical monocytes have a gene expression signature promoting resolution of inflammation and tissue repair

Gene expression indicates differential capabilities for antigen presentation, co-stimulation, and modulation of T-cell responses

Expression of metabolic genes differs markedly between classical and nonclassical monocytes

Mitochondrial respiration prevails in nonclassical monocytes

Gene set enrichment analysis

Single-cell transcriptomic data suggests continuous differentiation

Discussion

Data availability statement

Ethics statement

Author contributions

Funding

Acknowledgments

Conflict of interest

Publisher&#x;s note

Supplementary material

References

MyDdependent chemokine expression in mTECs^high

TLR9/MyD88 signaling in mTECs targets Sirpα⁺ cDC2

TLR9/MyD88 signaling in mTECs recruits CD14⁺moDCs