PDGFRα-lineage origin directs monocytes to trafficking proficiency to support peripheral immunity.

Multiple embryonic precursors give rise to leukocytes in adults while the lineage-based functional impacts are underappreciated. Mesodermal precursors expressing PDGFRα appear transiently during E7.5-8.5 descend to a subset of Lin- Sca1+ Kit+ hematopoietic progenitors found in adult BM. By analyzing a PDGFRα-lineage tracing mouse line, we here report that PDGFRα-lineage BM F4/80+ SSClo monocytes/macrophages are solely Ly6C+ LFA-1hi Mac-1hi monocytes enriched on the abluminal sinusoidal endothelium while Ly6C- LFA-1lo Mac-1lo macrophages are mostly from non-PDGFRα-lineage in vivo. Monocytes with stronger integrin profiles outcompete macrophages for adhesion on an endothelial monolayer or surfaces coated with ICAM-1-Fc or VCAM-1-Fc. Egress of PDGFRα-lineage-rich monocytes and subsequent differentiation to peripheral macrophages spatially segregates them from non-PDGFRα-lineage BM-resident macrophages and allows functional specialization since macrophages derived from these egressing monocytes differ in morphology, phenotype, and functionality from BM-resident macrophages in culture. Extravasation preference for blood PDGFRα-lineage monocytes varies by tissues and governs the local lineage composition of macrophages. More PDGFRα-lineage classical monocytes infiltrated into skin and colon but not into peritoneum. Accordingly, transcriptomic analytics indicated augmented inflammatory cascades in dermatitis skin of BM-chimeric mice harbouring only PDGFRα-lineage leukocytes. Thus, the PDGFRα-lineage origin biasedly generates monocytes predestined for BM exit to support peripheral immunity following extravasation and macrophage differentiation.

Introduction

Endothelium acts as the gate keeper to control leukocyte trafficking to and from circulation. A low level of trafficking is allowed under homeostasis for leukocyte turnover and immune surveillance. This gateway widens in inflamed tissues as the local endothelium becomes activated and expresses surface molecules to facilitate leukocyte recruitment [1]. Capture of fast‐flowing leukocytes in the circulation into rolling cells on endothelium requires interactions with endothelial selectins [2]. Together with chemokines and additional lectin interactions, leukocyte integrins are activated by inside‐out signaling to increase adhesiveness which eventually decelerates the rolling to full arrest on endothelial cell (EC) adhesion molecules (CAMs) [3, 4, 5]. Integrin signals then propagate to enable leukocyte crawling to an appropriate exit site for transendothelial diapedesis [6, 7].

Monocyte is a cardinal component of the innate immunity derived from hematopoietic stem and progenitor cells (HSPC) via definitive hematopoiesis in BM. HSPC give rise to common myeloid progenitor which then sequentially produces macrophage and DC progenitor (MDP) and a monocyte/macrophage lineage‐committed common monocyte progenitor (cMoP) [8, 9]. The heterogenous monocyte population contains Ly6C+ classical and Ly6C‐ nonclassical cells. Ly6C‐ monocytes in blood potentially originate from Ly6C+ cells [10, 11, 12]. Monocyte egress from BM into circulation requires transendothelial migration controlled by the CCL2‐CCR2 chemotactic axis [13]. Although sinusoidal endothelium in BM is fenestrated, the fenestrae are impairment to cellular entities [14]. In addition, diapedesis across basement membrane might not be necessary as the basement membrane on the abluminal side of endothelium is discontinuous [15].

Upon inflammation, monocytes follow the typical extravasation cascade to assess inflamed tissues with an exception of certain nonclassical monocytes which constantly crawl on the luminal surface of the endothelium [16]. Extravasated monocytes could differentiate into DCs or macrophages and have extended impacts on inflammation. Contrasting the homeostatic nature of tissue‐resident macrophages (trMF) in steady state, monocyte‐derived macrophages are generally inflammatory. Since macrophage is phenotypically plastic and adapts to the surrounding microenvironment [17], trMF presumably support inflammation as well amidst the proinflammatory neighborhood. Rendezvoused pool of macrophages performs various functions including antigen presentation, cytokine production, clearance of pathogens, and apoptotic cells and efferocytosis [18].

The functional correlation of monocyte/macrophage ontology is unclear. trMF could originate from embryonic precursors or monocytes depending on the tissue. Alveolar macrophage, an embryonic‐derived and self‐sustained macrophage [19], when irradiated and replaced by BM‐derived macrophages, adopted almost an identical enhancer profile as the native alveolar macrophages [17], suggesting the cellular plasticity might have overridden any ontology‐driven difference. Unlike Flk+, mesodermal hemogenic endothelium, which essentially lineage to all hematopoietic cells in adults via a process called endothelial‐to‐hematopoietic transition [20], mesodermal precursors expressing PDGFRα (Pa) which temporarily emerge within a narrow window between E7.5 and E8.5 only give rise to a subset of hematopoietic cells [21]. In adults, Pa is considered as a marker for mesenchymal cells [22], a lineage distinct from either endothelial or hematopoietic cells. We, hence, hypothesize monocytes and macrophages derived from the Pa+ embryonic precursors might inherit lineage‐based functionality difference from the rest.

Here, we have analyzed whether monocytes and macrophages derived from Pa+ precursors present functional difference in vivo from nonlineage counterparts. Using lineage tracing mice labeling progeny cells of PDGFRα‐lineage with tdTomato, we demonstrate that the PDGFRα‐lineage cells in BM F4/80+SSClo monocytes/macrophages are mostly egressing Ly6C+ monocytes. These cells show stronger integrin expression and are more adhesive on endothelial CAMs than BM macrophages which are mostly non‐PDGFRα‐lineage. Egressing LFA‐1hiMac‐1hi monocytes, enriched for PDGFRα‐lineage cells, generate distinct macrophages from BM‐resident macrophages. In addition, we profiled the extravasation preference of PDGFRα‐lineage monocytes and found classical monocytes preferentially enter skin and colon and subsequently modify the resident pool of macrophages. Finally, with BM transplantation and transcriptomic analytics, we show that the extravasation ease of PDGFRα‐lineage leukocytes aggravates the inflammatory responses in the dermal stroma during atopic dermatitis (AD).

Results

PDGFRα‐lineage contributes to peripheral monocytes but little to BM‐resident macrophages

We investigated the contribution of PDGFRα‐lineage to HSPC and differentiated leukocytes in BMs using PDGFRα‐lineage tracing mice, where Pdgfra promoter on a BAC clone drives the expression of Cre‐recombinase that subsequently labels progeny cells with tdTomato. Lineage‐ Sca1+ Kit+ (LSK) cells contain HSC and multipotent progenitors, which give rise to LS‐K and LSK‐ oligopotent or lineage‐restricted progenitors. Observed in multiple bone types, these three upstream progenitor populations contain 27.6‐33.8% cells of PDGFRα‐lineage (Figure 1A‐C and Supporting information Figure S1). Our measurements align with a previous study which pulse traced the descendant cells of PDGFRα‐expressing precursors by tamoxifen‐inducible Cre‐recombinase at E8.0 and labeled 25% of LSK cells in fetal liver by E15.5.21 Nevertheless, the Pa(lin)+ fraction falls to an average of 13.4% (±0.4%) in F4/80+SSClo monocytes/macrophage compartment. Using EdU to pulse‐label freshly synthesized cells from proliferating progenitors, we verified the PDGFRα‐lineage contribution has been reduced since biogenesis of BM F4/80+SSClo monocytes/macrophage (Figure 1D). Within 4‐h, there tends to be more PDGFRα‐nonlineage F4/80+SSClo being labeled by EdU (+27.1% [±4.5%]) suggesting PDGFRα‐lineage cells might be generated at a slower rate than nonlineage counterparts (Figure 1D). This decline in PDGFRα‐lineage contribution could also be observed, but to, much milder extent, in BM CD115+ monocytes (21.8% (±1.2%)), neutrophils (24.2% (±1.4%)), and B cells (17.1‐22.5% depending on the maturation stages) (Figure 1A‐C and Supporting information Figure S2). Only slight declines in PDGFRα‐lineage contribution were observed in downstream progenitors of monocytes and neutrophils, Kit+Ly6C+PECAM‐1+ cMoP, and the Kit+Ly6C+PECAM‐1‐ neutrophil progenitors [23], which respectively harbor 25.7% (±1.1%) and 23.4% (±1.6%) cells from PDGFRα‐lineage (Figure 1A‐C and Supporting information Figure S1). These PDGFRα‐lineage hematopoietic cells are solely descended from embryonic, but not from adults, PDGFRα+ precursors, as we detected are <1% PDGFRα‐expressing cells among LSK, myeloid progenitors, F4/80+, Ly6G+, or CD19+ compartments in adult BM using a knock‐in reporter mouse line [22], where H2B‐GFP expression is driven by endogenous Pdgfra promoter in one allele (Supporting information Figure S3).

Figure 1

F4/80. (A‐C) Bone marrow (BM) cells of vertebra, femur, and tibia were isolated from PDGFRα‐lineage tracing mice and the contribution of PDGFRα lineage was measured by flow cytometry. In BM, LSK progenitors, neutrophil progenitors, common monocyte progenitors (n = 3 mice each, 3 batches), CD19+ B cells (CD23–CD24+IgM– pre‐/pro‐B, CD23–CD24+IgM+ immature B and CD23+CD24+ transitional B, n = 4–6 mice, 3 batches), Ly6G+ neutrophils (n = 4 mice, 4 batches), CD115+SSClo monocytes (n = 4 mice, 1 batches), and F4/80+SSClo monocytes/macrophages (n = 5 mice, 3 batches) were analyzed. Samples in (A‐C) were compared to LSK cells. (D) Four hours after intraperitoneal administration of 1 mg EdU, PDGFRα‐lineage contribution in freshly generated EdU+ BM monocytes/macrophages and EdU‐labelling efficiency in Pa(lin)+/– F4/80+SSClo were measured (n = 7 mice, 2 batches). (E) PDGFRα‐lineage contribution in F4/80+SSClo Ly6C+/– PB monocytes (n = 13 mice, 6 batches), CD115+SSClo Ly6C+/– PB monocytes (n = 4 mice each, 1 batch), neutrophils and B cells (n = 3 mice each, 1 batch each) were compared to BM counterparts average of all bone types measured in (A‐C). (F) Leukocyte subset composition in Pa(lin)+ and Pa(lin)– compartments of total cells (upper) or F4/80+SSClo cells (lower) in BM and PB were compared within individual mice (n numbers indicated in brackets). (G) Proportion of CX3CR1+ (n = 3 mice, 1 batch) or MHCII+ (n = 3 mice, 2 batches) cells in Pa(lin)+ and Pa(lin)‐ F4/80+SSClo monocytes/macrophages were quantified within individual mice. Error bars, SEM. Individual data are represented by dots. For all measurement, one mouse per experiment. Groups were compared by one‐way ANOVA followed by Tukey's multiple comparisons in (A‐C and E), two‐sample t‐test in (D) comparing Pa(lin)+ proportion and paired t‐test in (D) comparing EdU labeling between Pa(lin)+/– within individual mice and in (F‐G). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Since F4/80+SSClo BM cells contain a mixed population of monocytes and macrophages, we examined the CD115+ monocyte content within this population and the expression of F4/80 among the whole CD115+ monocyte population. In BM, F4/80+SSClo BM cells contains only 5.8% (±0.4%) CD115+ monocytes. Nevertheless, these F4/80+CD115+ monocytes are the major trafficking fraction as F4/80+SSClo cells in peripheral blood (PB) contains 90.1% (±0.4%) CD115+ cells. F4/80 expression marks 29.3% (±1.9%) of BM CD115+ monocytes with 25.3% (±1.7%) and 72.3% (±1.1%) of Ly6C+ classical and Ly6C– nonclassical monocytes (cMO and ncMO), respectively, being labeled. In PB, intermediate F4/80 expression could be detected on both cMO and ncMO (Supporting information Figures S4 and S5). Hence, BM F4/80+SSClo cells contain macrophages and a subset of CD115+ monocytes.

Within the BM F4/80+SSClo monocytes/macrophages population, we found that the PDGFRα‐lineage contribution deficiency is specific to the Ly6C– compartment. The fact that F4/80+Ly6C– cells receive a significantly lower PDGFRα‐lineage contribution than CD115+Ly6C+/– monocytes indicates macrophages in BM being mostly derived from non‐PDGFRα‐lineage monocytes (Figure 1E). On the other hand, both PDGFRα‐lineage and nonlineage contribute to BM F4/80+Ly6C+ cells at a similar ratio to CD115+Ly6C+ monocytes (Figure 1E). Previous BrdU pulse‐trace experiments have shown CCR2+Ly6C+ cMO as the monocyte subset that readily egresses to circulation [10, 11] and gives rise to blood Ly6C– ncMO in situ [10]. Egress efficiency of these cMO, either F4/80+ or CD115+, are equivalent between PDGFRα‐lineage/nonlineage as the Pa(lin)+ fraction do not differ between BM and PB. Similarly, egress efficiency of Ly6G+ neutrophils and CD19+ B cells do not differ by PDGFRα‐lineage (Figure 1E). In general, PDGFRα‐lineage BM cells contain more CD115+ monocytes and Ly6G+ neutrophils (Figure 1F, upper panel). Among BM F4/80+SSClo monocytes/macrophages, PDGFRα‐lineage is enriched for egressing CD115+ and Ly6C+ cells (Figure 1F, lower panel and Supporting information Figures S4 and S5) but not for CX3CR1+ or MHCII+ cells (Figure 1G). Taken together, while the non‐PDGFRα‐lineage F4/80+SSClo encompasses both egressing monocytes and BM‐resident monocytes that differentiate to macrophages, the PDGFRα‐lineage mostly generates Ly6C+ egressing monocytes.

PDGFRα‐lineage F4/80+ monocytes are enriched on the abluminal surface of femoral endothelium

Sca‐1–PECAM‐1+ sinusoids have been reported as major trafficking sites [24]. If PDGFRα‐lineage cells are restricted to the trafficking monocytes, we hypothesize these cells might populate the surface of these trafficking portals. Three‐dimensional confocal examination of adherent F4/80+ cells (in directly contact) on the sinusoidal abluminal surface revealed 36.7% (±3.6%) of the cells observed were from PDGFRα‐lineage, a significantly higher proportion than the global Pa(lin)+ fraction of F4/80+ monocytes/macrophages in the whole femur as measured by flow cytometer (Figure 2A and B). Enriched Pa(lin)+ F4/80+ cells on the abluminal endothelium suggest BM monocytes, but not macrophages, are in close association with these sinusoidal trafficking portals.

Figure 2

The abluminal surface of sinusoidal endothelium are enriched for PDGFRα‐lineage F4/80. (A) Spatial correlation between PDGFRα‐lineage monocytes and sinusoidal abluminal surface was examined in 60 μm‐thick BM cryosection by three‐dimensional confocal microscopy. A representative image of five experiments (upper panel) is shown as maximal intensity projection (MIP). Scale, 30 μm. Inset was shown as a single slice (lower panel). Left arrow indicates a Pa(lin)+ monocyte and right arrow indicates a Pa(lin)‐ monocytes. Scale, 10 μm. (B) Proportion of Pa(lin)+ monocytes on sinusoids (n = 177 monocytes on 34 sinusoids pooled from five mice) was quantified and compared to the global proportion of Pa(lin)+ monocytes in femur BM (n = 16 mice, 6 batches) measured by flow cytometry. Error bars, SEM. Individual data are represented by dots. One mouse per experiment. Images are representative of five independent experiments. Groups were compared by Mann–Whitney U‐test. ****p < 0.0001.

Integrins favor monocyte over macrophage adhesion on ECs or CAMs in vitro

We tested whether PDGFRα‐lineage‐rich monocyte fraction of F4/80+SSClo preferentially adheres on ECs, a necessary step in transendothelial migration. Bone marrow cells (BMC) obtained from the PDGFRα‐lineage tracing mice were allowed to adhere on resting or TNF‐α stimulated primary EC isolated from WT skin in Mg2+ and Ca2+ containing serum‐free medium. After dissociating adherent cells from the endothelial monolayer, the Pa(lin)+ fraction among adherent F4/80+SSClo was measured and compared to the input. Since Pa(lin)+ cells in F4/80+SSClo are restricted to monocytes (Figure 1E), an increased Pa(lin)+ fraction after adhesion indicates monocytes better compete over Pa(lin)–rich macrophages for the adherent surface. After adhesion on resting EC, we recorded a significant increase in the Pa(lin)+ fraction of F4/80+SSClo, ranging from +50.3% (±7.2%) to +80.0% (±2.6%). TNF‐α‐stimulated EC supported at least fivefold more adhesion than at rest and similarly, stronger adhesion for PDGFRα‐lineage‐rich F4/80+SSClo monocytes was observed with similar magnitude ranges from +41.7% (±3.7%) to +71.2% (±3.0%) (Figure 3A and B).

Figure 3

Altered integrin expression and activation support better adhesion of PDGFRα‐lineage F4/80. (A‐B) Bone marrow cells containing both PDGFRα‐lineage and nonlineage BM cells were allowed to competitively adhere on resting (blue) or TNF‐α inflamed (red) primary dermal endothelial cells. (A) Adherent F4/80+SSClo were dislodged and enumerated. (B) Proportion of PDGFRα lineage in adherent F4/80+SSClo were compared to that in input cells (black). A total of n = 10 each from two independent experiments for resting conditions and n = 18 each from three independent experiments for TNF‐α stimulation. (C‐D) The competitive adhesion assay described in (A‐B) was repeated on surfaces coated with ICAM‐1‐Fc (blue) or VCAM‐1‐Fc (red). n = 10 each from two independent experiments. Adherent cells were quantified in (C) and the proportion of Pa(lin)+ F4/80+SSClo in adherent cells were compared to input in (D). (E) Experiment described in (C‐D) was repeated using either CCL2 or MnCl2 (n = 16 and 8 pooled from two independent experiments) as the stimulus. (F‐G) Expression of LFA‐1, integrin α4 and integrin β1 on surfaces of Pa(lin)+ and Pa(lin)‐ F4/80+SSClo from indicated tissues and subsets were measured by flow cytometry. n = 3 mice in three batches for (F), n = 3 mice in one batch for PB in (G) and n = 5–6 mice in two batches for BM in (G), one mouse per experiment. (H) Experiment described in (E) was repeated with PB cells (n = 5 each in an experiment with cells pooled from four mice). Error bars, SEM. Individual data are represented by dots. Groups were compared by two‐sample t‐test in (A and C), one‐sample t‐test against input in (B and D) and paired t‐test in (F‐G). In (E), groups were compared against the CCL2 stimulation ctrls by one‐way ANOVA followed by Tukey's multiple comparisons and by one‐sample t‐test against input. #p < 0.05, ####p < 0.0001 against CCL2 stimulation. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Since adhesion on EC is solely supported by interaction with CAMs, we assessed if ICAM‐1 or VCAM‐1 alone, the two major endothelial CAMs, suffices to favor adhesion for PDGFRα‐lineage‐rich monocytes. The competitive adhesion assay was repeated on a surface immobilized with purified proteins of either ICAM‐1‐Fc or VCAM‐1‐Fc in the presence of CCL2 as the stimulating chemokine. Similar to the endothelial surface, adhesion on both immobilized ICAM‐1‐Fc and VCAM‐1‐Fc led to monocyte enrichment in magnitudes comparable to that on EC (Figure 3C and D). Leukocytes interact with endothelial CAMs via integrins which need to be activated into a high affinity and open conformation for ligand interaction. To delineate whether the better adhesion support for PDGFRα‐lineage‐rich monocytes is due to more efficient activation, in the competitive adhesion assay, we replaced the stimulating reagent of CCL2 with Mn2+, which bypasses the inside‐out signaling and forces integrins into high‐affinity open conformation [25, 26]. Mn2+ stimulation reduced the adhesion advantage of Pa(lin)+‐rich monocytes fully on ICAM‐1‐Fc and partially on VCAM‐1‐Fc (Figure 3E). The partial reduction of the adhesion advantage after bypassing the inside‐out signal suggests integrin heterodimer expression, besides activation, also contributes to the adhesion advantage of marrow monocytes over macrophages on endothelial CAMs.

LFA‐1 (ItgαL/β2) and VLA‐4 (Itgα4/β1) are the major integrins utilized by leukocytes to interact with ICAM‐1 and VCAM‐1, respectively [27]. To directly examine if integrin expression differs between PDGFRα‐lineage and between monocytes and macrophages, we measured the expression of LFA‐1, Itgα4, and Itgβ1 on the surface of marrow F4/80+SSClo monocytes/macrophages. While Pa(lin)+ monocytes express significantly more LFA‐1 than Pa(lin)– monocytes/macrophages, individual expression of Itgα4 and Itgβ1 do not massively differ (Figure 3F). Stronger LFA‐1 expression was also detected on F4/80+SSCloLy6C+ monocytes than on Ly6C– macrophages (Figure 3G). After entry to circulation, monocytes express high levels of integrin which do not differ much by PDGFRα‐lineage (Figure 3G). Accordingly, PB Pa(lin)+ monocytes do not preferentially adhere on the CAMs upon chemokine or Mn2+ stimulation (Figure 3H).

BM monocytes generate macrophages functionally distinct from marrow‐resident macrophages

When the expression of ItgαM, the integrin alpha chain of another ICAM‐1‐binding integrin Mac‐1, on BM F4/80+SSClo monocyte/macrophage was measured, we detected a high expression of ItgαM on PDGFRα‐lineage F4/80+SSClo monocytes. Indeed, Pa(lin)+ BM F4/80+SSClo monocytes are essentially LFA‐1hi ItgαM hi whereas Pa(lin)– BM F4/80+SSClo, comprising of monocytes and macrophages, contain both LFA‐1hi ItgαM hi and LFA‐1lo ItgαM lo cells (Figure 4A). Expression levels of LFA‐1 and ItgαM are similar between PDGFRα‐lineage/nonlineage LFA‐1hi ItgαM hi monocytes (only differing by 5.0 ± 1.0% and 6.3 ± 1.5% for LFA‐1 and ItgαM, respectively). PDGFRα‐lineage contribution in LFA‐1hi and/or ItgαM hi F4/80+SSClo are similar to that in F4/80+ or CD115+ Ly6C+ cMO in BM and PB, confirming this population contains mostly monocytes (Figures1E and 4B). In a pulse‐trace experiment by EdU to follow the kinetics of freshly generated LFA‐1hi ItgαM hi and LFA‐1lo ItgαM lo F4/80+SSClo (Mhi2 and Mlo2), we found that at 4 h post‐ EdU administration, EdU+ cells could be detected among both Mhi2 and Mlo2 with the labeling in Mlo2 overwhelming. Frequency of EdU+ Mlo2 declined after 1 day accompanying increased frequency of EdU+ Mhi2 (Figure 4C). This temporal sequence of EdU labeling suggests Mlo2 might contains some monocytes, mostly from non‐PDGFRα‐lineage, that have turned LFA‐1hi ItgαM hi. Direct EdU tracing of non‐PDGFRα‐lineage F4/80+SSClo cells confirmed this conversion (Supporting information Figure S6).

Figure 4

Distinct expression of LFA‐1 and Mac‐1 support egress of PDGFRα‐lineage F4/80. (A) BM cells were isolated from PDGFRα‐lineage tracing mice and measured for expression of LFA‐1 and ItgαM. (B) Fraction of Pa(lin)+ cells was measured in all F4/80+SSClo cells, in the LFA‐1hi, ItgαM hi, or LFA‐1hi ItgαM hi fractions. n = 12 mice, four batches. (C) Four hours or 10 day (n = 5 or 4 mice pooled from two batches) after intraperitoneal administration of 1 mg EdU to WT mice, frequencies of EdU+ BM monocytes in the indicated subsets were measured. (D) LFA‐1hi ItgαM hi and LFA‐1lo ItgαM lo F4/80+SSClo monocytes were isolated from WT and differentiated in M‐CSF containing medium for 7 days. Expression of indicated macrophage markers were measured by flow cytometry. n = 6 differentiation from three mice (two differentiation per experiment, three experiments). (E) Morphology of the derived macrophagesis shown. Scale, 100 μm (upper) and 50 μm (inset). Images are representative of four experiments. (F‐H) Macrophages were derived from LFA‐1hi ItgαM hi and LFA‐1hi ItgαM hi F4/80+SSClo monocytes in vitro and analyzed by bulk RNA sequencing. (F) Significant differentially expressed genes were analyzed for functional enrichment (Gene Ontology Biological Process). Top five enriched functions are shown for each macrophages. (G) Transcription factor (TF) activities were measured for each macrophages. TF activities with >2 SD difference are shown. (H) Common markers and surface receptor markers distinguishing Machi2 and Maclo2 are shown. n = 2 differentiation pooled per mouse, three mice in total for bulk RNA‐seq. (I) F4/80+Ly6C–CD64+ WT BM macrophages were examined for expression distribution of Siglec‐1 and CXCR4. n = 3 mice, one batch. (J) PDGFRα‐lineage contribution to Siglec‐1+CXCR4– and Siglec‐1–CXCR4+ cells among F4/80+Ly6C–CD64+ BM macrophages isolated from lineage tracing mice were measured. n = 7 mice, four batches. Error bars, SEM. Individual data are represented by dots. For (A‐C) and (I‐J), one mouse per experiment. Groups were compared by two‐sample t‐test in (A, ,D and I‐J) and one‐way ANOVA followed by Tukey's multiple comparisons in (B‐C). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Macrophages derived in vitro from Mhi2 and Mlo2 (Machi2 and Maclo2), respectively, corresponding to peripheral macrophages differentiated from trafficking monocytes and BM‐resident macrophages, expressed similar amount of ItgαM. CD64 expression was higher on Machi2 (Figure 4D). Maclo2 tended to be more roundish whereas Machi2 were more elongated and spikier (Figure 4E). Macrophage differentiation efficiency is unaffected by lineage origin as Pa(lin)+ proportion in F4/80+ cells remained unaltered after differentiation (Supporting information Figure S7). We explored by bulk RNA sequencing whether Machi2 and Maclo2 might perform different functions. In these macrophages, expression of 3949 genes significantly differed by the originating F4/80+SSClo. Functional enrichment analyses of these genes indicated Machi2 differs from Maclo2 by stronger expression of cell cycle regulating genes whereas gene expression in Maclo2 tilts toward immune responses (Figure 4F). Machi2 is characterized by high transcription factor (TF) activities of cell cycle relevant TF such as E2F family, forkhead box protein M1 and Myc. In contrast, higher TF activities relevant to cell activation, such as STAT1/3, PU1, and NF‐kB, were detected in Maclo2 (Figure 4G). Transcript expression of common macrophage markers did not differ between Machi2 and Maclo2. Interestingly, Machi2 expresses high levels of Siglec‐1 and Cx3cr1, both of which are known markers of marrow‐resident macrophages, suggesting a portion of Mhi2 has remained in BM and given rise to Machi2. In contrast, Maclo2 expresses low transcript levels of Siglec‐1 and Cx3cr1 but high levels of Cxcr4 and potentially represent another macrophage subset in BM (Figure 4H). In vivo, a limited fraction of F4/80+Ly6C–CD64+Siglec‐1+ BM macrophages express CXCR4 while most F4/80+Ly6C–CD64+Siglec‐1– cells are CXCR4+ (Figure 4I). When we re‐examined our previous single‐cell RNA sequencing datasets (GSE159535) and mapped PB monocytes to BM monocytes/macrophages, PB monocytes overlay with cluster 3 and 1 of BM monocytes/macrophages, demonstrating cluster 3 and 1, but not cluster 2, of BM monocytes/macrophages are sources of egressed monocytes (Supporting information Figure S8A). Meanwhile, Siglec‐1+ macrophages were exclusively found in cluster 1 and, thus, are likely derived from egress‐ready Mhi2‐like monocytes. Stronger Cxcr4 signal was detected in cluster 2 and these macrophages are derived from BM‐resident Mlo2‐like monocytes (Supporting information Figure S8B). Accordingly, Machi2‐like F4/80+Ly6C–CD64+Siglec‐1+CXCR4– has more PDGFRα‐lineage descendants than Maclo2‐like F4/80+Ly6C–CD64+Siglec‐1–CXCR4+ macrophages (Figure 4J). Overall, we show the Pa(lin)+‐rich BM egressing monocytes differentiate to macrophages that differ in functionality from Pa(lin)– BM‐resident macrophages.

Extravasation preference for PDGFRα‐lineage monocytes is controlled by tissue‐specific determinants

We next asked if extravasation of monocytes differs by PDGFRα‐lineage origin. We have here examined in vivo the competitive extravasation of PDGFRα‐lineage/nonlineage monocyte to tissues (skin, peritoneum, and colon) under steady state and inflammation.

Local acute inflammation was induced intradermally by injecting 1 μg LPS or saline ctrl to different dorsal areas of the same mouse. Skin samples were collected overnight and analyzed by flow cytometry. Comparing the two injection areas verified a LPS‐specific inflammation with a 5.6‐fold (±2.1‐fold) rise in the infiltrated F4/80+ MHCII‐ Ly6C+ cMO in the F4/80+ compartment (Figure 5A). To determine if extravasation efficiency of monocytes differs by PDGFRα‐lineage origin, we compared the Pa(lin)+ fractions in pre‐extravasated monocytes in the circulation and postextravasated monocytes in LPS‐stimulated skin. To our surprise, in contrast to 21.6% (±2.1%) of Pa(lin)+ monocytes in the blood, the Pa(lin)+ fraction expanded by 57% after extravasation to the inflamed skin (Figure 5B). Interestingly, this effect was specifically observed with cMO (Figure 5B and C).

Figure 5

Recruitment preference for PDGFRα−lineage monocytes influences the lineage composition of tissue macrophages and is dictated by tissue‐specific determinants under steady state and inflammation. (A‐C) Mice received a single dose of LPS and saline ctrl on distinct dorsal skin areas. After overnight, skin was excised and analyzed for monocyte infiltration by flow cytometry. (A) Monocyte infiltration to LPS‐stimulated area and ctrl area relative to F4/80+ cells (left) or all cells (right) were measured. Proportion of Pa(lin)+ classical monocytes (B) and nonclassical monocytes (C) in peripheral blood and inflamed skin are shown. n = 3 mice each, two induction batches. (D) Six bi‐/tridaily doses of MC903 were applied on dorsal skin and inflammation was monitored across 14‐day time course. n = 10 mice, three induction batches. Inflamed skin was excised and analyzed on day 14 by flow cytometry. (E) Monocyte infiltration was analyzed (n = 4 [healthy, 2 batches] or 5 [AD, 2 induction batches]). Proportion of Pa(lin)+ classical monocytes (F) and nonclassical monocytes (G) were compared between peripheral blood and inflamed skin. n = 5 mice each, two induction batches. (H) Proportion of Pa(lin)+ macrophages in healthy (n = 8 mice, 5 batches), LPS‐inflamed (n = 3 mice in [A‐C]) and atopic dermatitis (n = 5 mice in [E‐G]) skin are shown. Proportion of (I) F4/80hi and F4/80+ cells, and (J) the respective Ly6C+/– subsets among F4/80+SSClo monocyte/macrophage in peritoneal cavity (PC) were measured under indicated conditions. (K) PDGFRα‐lineage contribution to the indicated subsets was measured under TNF‐α stimulation. Reference measurements under steady state are indicated. n = 4 (PB, three induction batches) and five mice (each for inflamed/untreated peritoneal cavity, three induction batches). (L) PDGFRα‐lineage contribution in the indicated colonic monocyte/macrophage subsets were measured and compared to the peripheral blood (PB) average measured in Figure 1E (purple dashed line). n = 9 mice, three batches. Error bars, SEM. Individual data are represented by dots. For comparisons of PDGFRα‐lineage contribution between tissue compartments or frequencies between cellular subsets in (B‐C, F‐G, and J‐L), one mouse per experiment. Groups were compared by two‐tailed t‐test in (A‐F and I‐J), one‐way ANOVA followed by Tukey's multiple comparisons in (G and K) and one‐sample t‐test in (L). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

To trigger chronic inflammation in the AD model, we applied 5 nmol MC903 which stimulates production of TSLP from keratinocytes [28] on dorsal skin bidaily or tridaily for six times. Thickened and/or reddened inflamed area was monitored in time course. The MC903 treatment induced a progressive inflammation that peaked on day 11 (Figure 5D). Monocyte influx into the inflamed skin was verified (Figure 5E) and extravasation efficiency of PDGFRα‐lineage monocytes was examined on day 14. Similar to LPS stimulation, enhanced extravasation was observed with PDGFRα‐lineage cMO. Pa(lin)+ cells constituted 24.1% (±1.2%) of the pre‐extravasated PB monocytes and 37.6% (±3.6%) in the postextravasated population in the skin (Figure 5F and Supporting information Figure S9). Extravasation of ncMO was unaffected (Figure 5G). In addition, no specific expansion of PDGFRα‐lineage monocytes in BM was evident (Supporting information Figure S10). We found that F4/80+ MHCII+ dermal macrophages in both healthy and AD skin receive major PDGFRα‐lineage contribution (Figure 5H and Supporting information Figure S11). Using 3D‐confocal microscopy to locate large and irregular‐shaped macrophages in AD skin, we observed an agreeing proportion of 38% of macrophages being from PDGFRα‐lineage (Supporting information Figure S12). Since dermal macrophages are continuously replenished by extravasated monocytes even under homeostasis [29], it suggests that the selective extravasation of PDGFRα‐lineage cMO has shaped the lineage composition in dermal macrophages. When the AD macrophage composition was analyzed by single‐cell RNA‐sequencing, three transcriptomic subtypes differing in surface markers, cytokine expression, and transcription activities, were identified. PDGFRα‐lineage macrophages manifested a higher ratio of subtype 1/2 than nonlineage (3.7 vs 1.4) (Supporting information Figure S13), implying the lineage origin could have altered the functional profile of the descendant macrophages.

The extravasation support to PDGFRα‐lineage cMO and the subsequent macrophage formulating capability seems to be tissue‐specific. In unstimulated peritoneal cavity (PC), there is a major F4/80hi Ly6C–, a minor F4/80+ Ly6C– macrophage populations, and ∼5% F4/80+ Ly6C+ infiltrating cMO. TNF‐α stimulation specifically expanded the relative proportion of Ly6C+ infiltrating cMO (Figure 5I and J and Supporting information Figure S14). PDGFRα‐lineage contribution in F4/80hi macrophages was slightly lower than that in infiltrating cMO, while equivalent PDGFRα‐lineage contribution was observed in F4/80+ macrophages and cMO, suggesting F4/80+ macrophages being derived from infiltrating monocytes. Infiltrating cMO specifically contributing to F4/80+ macrophages has been described with LPS or thioglycolate stimulation [30]. Nevertheless, PDGFRα‐lineage contribution in neither of these monocyte/macrophage compartments exceeded the level in blood, indicating PDGFRα‐lineage monocytes had not been favored for peritoneal extravasation (Figure 5K). We further examined whether preferential extravasation of PDGFRα‐lineage cMO could be observed in colon, yet another tissue receiving constant monocyte input [31]. Ly6C+ cMO constitutes 8.2% (±1.2%) of F4/80+SSClo cells in colon lamina propria under steady state. PDGFRα‐lineage contribution in these extravasated cMO (51.0% (±3.3%)) significantly exceeded that in blood, thus, indicating supported extravasation of PDGFRα‐lineage cMO. Similar enhancement could be observed in F4/80+MHCII+ macrophages (33.6% (±1.1%)) (Figure 5L and Supporting information Figure S14) or F4/80+MHCII+CX3CR1+ macrophages (33.9% (±1.8%)) (Supporting information Figure 15). Of note, PDGFRα‐lineage contribution to macrophages in all these peripheral tissues are prominently higher than that in BM (Figure 4J), highlighting the roles of PDGFRα‐lineage monocytes in peripheral defense. Hence, among the three examined tissues where at least a subset of macrophages receives turnover from infiltrating monocytes, we observed favored extravasation of PDGFRα‐lineage cMO in skin and colon but not in peritoneum. These extravasated monocytes subsequently modulate the lineage composition of tissue macrophages.

Extravasated PDGFRα‐lineage leukocytes differentially regulate the dermal environment

Since the extravasated pool of monocytes contained mixed PDGFRα‐lineage origins, it renders investigation of the impacts of postextravasated PDGFRα‐lineage monocytes to the tissue stromal environment difficult. To this end, we performed BM transplantation (BMT) to reconstitute the hematopoietic system of irradiated EGFP‐Tg recipients with either Pa(lin)+ or Pa(lin)– Lineage– progenitors so that subsequent extravasation would be limited to a single PDGFRα‐lineage origin. After 6‐week recovery, PB was examined and verified approximately 90% blood cells being donor derived (Figure 6A). To assess in AD how the dermal environment might be influenced by the extravasated leukocytes, which were of single PDGFRα‐lineage origin in these chimera mice, the mice were subjected to AD induction by MC903. Mice that had received Pa(lin)+ transplant presented more severe skin inflammation (1.98‐fold on average weighed by inflammation severity) (Figure 6B). On day 14, skin was sampled for bulk RNA‐sequencing and flow cytometric analysis of recruited leukocytes. Despite a lack of competitive control, the numbers of both recruited donor‐derived CD45+ leukocytes and CD45+ F4/80+ monocytes/macrophages normalized to recipient‐derived stromal cells tended to be higher in mice receiving Pa(lin)+ transplant (Figure 6C). In the sequencing analysis, 237 genes among the top half of expressed genes showed more than twofold expression difference between Pa(lin)+ and Pa(lin)– BMT (Figure 6D). Gene ontology (GO) analysis of these genes indicated enriched inflammatory functions accompanying reduced dermal homeostatic functions with Pa(lin)+ BMT (Figure 6E). These results correlate well with the extravasation proficiency of PDGFRα‐lineage monocytes. We sought to understand whether these affected processes by Pa(lin)+ BMT might associate with leukocyte PDGFRα‐lineage origin independently of the extravasation effect. Leukocyte recruitment varied among the sequenced skin samples and fell into two sample clusters. All except one Pa(lin)– BMT samples were found in the weak recruitment cluster whereas Pa(lin)+ BMT samples were divided (Figure 6F). Linear regressions were performed to investigate the independent contribution of extravasation and PDGFRα‐lineage origin of the transplant to the top 10 enriched or depleted GO biological processes (GOBP) (see Materials and Methods for details). GOBP affected by Pa(lin)+ BMT are solely associated with extravasation (alone or interaction with BMT PDGFRα‐lineage origin) but not to leukocyte PDGFRα‐lineage origin alone (Figure 6G and Supporting information Table S1 and S2). Thus, the impacts to the dermal stromal environment brought about by PDGFRα‐lineage leukocytes are large via extravasation. In addition, GSEA analysis [32], which relies on ranked expression fold ratio and avoid manual thresholding, reported consistent results with GO analyses and additionally indicated matrix metalloproteinases activation and ECM remodeling in Pa(lin)+ BMT (Supporting information Table S3).

Figure 6

Leukocytes of PDGFRα‐lineage origin exert differential influence to the global dermal transcriptome in atopic dermatitis via extravasation. Irradiated EGFP‐Tg mice received BM transplantation (BMT) of lin‐ Pa(lin)+ or Pa(lin)– progenitors. After reconstitution for 6 weeks, PB was sampled and analyzed for transplantation efficiency and specificity. (A) Representative flow cytometry plots of mice receiving Pa(lin)+ (left) or Pa(lin)– (middle) BMT and quantification (right) are shown. (B) AD was induced by MC903 and inflamed area was monitored during the time course. (C) Leukocytes infiltrated to AD skin were analyzed by flow cytometry on day 14. n = 4 Pa(lin)+ and 7 Pa(lin)– BMT‐chimeric mice in an experiment. Error bars, SEM. Groups were compared by two‐way ANOVA in (B) and two‐tail t‐test in (C). (D) RNA was purified from AD skin of these mice on day 14 postinduction and processed for bulk RNA‐sequencing. Top 50% expressing genes showing >twofold change in expression between BMT PDGFRα‐lineage origins are shown. (E) Genes in (D) were analyzed for functional enrichment. Top five significant enriched GO biological processes are shown. (F‐G) Linear regression was performed with top 10 each of positively or negatively regulated GOBP. (F) Extravasation indices were computed from flow cytometric measurement of leukocyte recruitment in sequenced samples (W: weak, S: strong recruitment group). (G) Aggregated expression of enriched genes in a concerned GOBP was regressed against the extravasation index and PDGFRα‐lineage origin of BMT. Estimates of coefficient a, b, and c (associated variable given in [F]) are shown as heatmaps with a p‐value cutoff at 0.05.

Discussion

Leukocytes in adult mice lineage are from a cocktail of embryonic precursors [33]. Among these precursors are mesodermal embryonic progenitors, transiently expressing PDGFRα between E7.5 and E8.5, which give rise to PDGFRα‐lineage leukocytes [21]. While previous simple transcriptomic comparison of PDGFRα‐lineage/nonlineage CD45+ leukocytes found overall similarity [34], it had not assessed functional aspects, in particular, those swift responses solely regulated by post‐transcriptional mechanisms, of specific immune compartments. Here, by following the trafficking dynamics of F4/80+SSClo monocytes/macrophages from BM to tissues, we reveal different functional fates of monocytes by PDGFRα‐lineage origin. F4/80+SSClo monocytes/macrophages differentiation from upstream hematopoietic progenitors accompanies a dwarfed PDGFRα‐lineage compartment in BM. This effect is due to BM‐resident macrophages, characterized as Ly6C–Siglec‐1–CXCR4+, being mostly derived from non‐PDGFRα‐lineage monocytes. PDGFRα‐lineage monocytes specifically occupy the LFA‐1hi ItgαM hi compartment, within which CD115+Ly6C+ monocytes dock on the abluminal perivascular surface of BM sinusoids and egress to peripheries. Upon extravasation to some tissues, local recruitment cues favor recruitment of PDGFRα‐lineage PB monocytes, which accordingly formulate the tissue macrophage lineage composition (Figure 7).

Figure 7

Proposed model showing monocyte fates from biogenesis to trafficking to tissues by PDGFRα‐lineage origin. After differentiation from hematopoietic progenitors, monocytes either intravasate to circulation or differentiate to macrophages in BM. BM‐resident macrophages are solely derived from non‐PDGFRα lineage while both lineages supply trafficking monocytes to circulation as LFA‐1hiMac‐1hi CD115+Ly6C+ cells. Depending on the tissue‐specific recruitment cues, PDGFRα‐lineage classical monocytes preferentially extravasate to tissues.

Compared to LSK progenitors, BM leukocytes, in general, tend to receive reduced contribution from PDGFRα‐lineage. In our EdU pulse‐trace experiment which labeled newly formed EdU+ BM F4/80+SSClo monocytes/macrophages, PDGFRα‐lineage cells seem to be formed slower than non‐lineage. It, therefore, implies reluctance of PDGFRα‐lineage progenitors to differentiate, likely a gradual process as we detected intermediate PDGFRα‐lineage contribution in committed progenitor cMoP and monocytes. Similar reduction in PDGFRα‐lineage contribution could be observed in BM B cells. In fetal liver, however, B cells receive similar PDGFRα‐lineage contribution as LSK cells [21] suggesting postnatal nature of differentiation bias. Since differentiation decisions are heavily influenced by genomic accessibility which precede transcriptional alteration [35], high‐throughput inspection of chromatin structure and accessibility of PDGFRα‐lineage/nonlineage LSK might provide insights on their differentiation potentials in vivo. We consider the alternative possibility that PDGFRα‐lineage BM leukocytes being short‐lived unlikely, since we observed fewer, but not more, PDGFRα‐lineage cells in the dead cell gate in our flow cytometric analysis of these BM leukocytes (not shown).

Endothelia are the major barriers controlling the leukocyte traffics across BM, circulation, and tissues. We show in BM that PDGFRα‐lineage F4/80+SSClo cells exist mostly as Ly6C+ LFA‐1hi ItgαM hi trafficking monocytes, but not macrophages. Monocytes display a stronger integrin‐based adhesive surface which under chemoattraction enables them being better picked up by the endothelium for release into the blood. Unlike PDGFRα‐lineage monocytes, non‐PDGFRα‐lineage monocytes participate in both BM macrophage differentiation and trafficking. Deriving macrophage cultures (Machi2 and Maclo2) from trafficking monocytes (LFA‐1hi ItgαM hi) or BM‐resident macrophages (LFA‐1lo ItgαM lo), our transcriptomic analysis has revealed different functional specialization, TF activities, and gene markers in these macrophages. As Machi2 is derived from trafficking monocytes, they most likely emerge after extravasation to tissues in vivo. Hence, Machi2 not only segregates from Maclo2 in location (peripheries vs BM) but the functions as well. Despite the ultimate molecular and functional landscape of macrophages being the subject of plastic remodeling by vicinity cues [17, 36], Machi2 expresses Cx and CX3CR1+ macrophages have been shown to be derived from circulating monocytes in peripheral tissues of, for example, colon [31] and liver [37]. Another study has shown that FoxM1 activity, strongly detected in Machi2, is required for generating CX3CR1+ pathological monocyte‐derived osteoclasts in arthritis [38]. On the other hand, Maclo2 differs from the well‐known Siglec‐1+ (which is expressed by Machi2) marrow macrophages [39] and represents another subset of marrow‐resident macrophages expressing CXCR4. In our single‐cell transcriptome dataset that maps BM monocytes/macrophages against PB monocytes (Supporting information Figure S8), while BM‐resident monocytes/macrophages express the strongest CXCR4 signal, lower level of CXCR4 expression could be detected in cells mapped to proximity of PB monocytes. This lower expression of CXCR4 seems to be coherent to a previous study [40] describing CXCR4 dowregulation during biogenesis of egressing monocytes.

We noticed in some tissues that extravasation of the PDGFRα‐lineage fraction had been favored. While the responsible tissue‐specific determinants remain unknown, wherever PDGFRα‐lineage monocytes are preferentially extravasated, to skin, for instance, the support tends to be more prominent with cMO. An unusual mode of extravasation has been described for peripheral ncMO that they constantly crawl on endothelium in a LFA‐1‐dependent manner even under steady state to allow quick response to invasion [16]. Likewise, a marginal pool of human CD16+ monocytes with high‐adhesion molecules expression has been reported [41]. During conversion of cMO to ncMO in PB [10], integrin expression on cell surface is upregulated and become available for activation and EC interaction to support the patrolling behavior observed under steady state [6, 7, 42]. Distinct from integrin‐equivalent PB Pa(lin)+/– cMO, Pa(lin)+ ncMO seems to express a lower amount of Itgβ1, which could pose an adhesive disadvantage and counteract potential recruitment advantage. It has been further suggested that ICAM‐1 density on the apical endothelial surface, and therefore, should be the leukocyte‐EC adhesive strength, influences the subsequent diapedesis route [43]. Hence, the differently regulated ncMO extravasation mechanism might partially account for the selective recruitment of PDGFRα‐lineage cMO to tissues. In addition, by mapping single‐cell RNA sequencing data of PB monocytes to BM monocytes/macrophages, we could identify subsets of BM monocytes/macrophages that egress to PB. Similar strategies could be pursued in the future to distinguish PB monocytes subsets infiltrating skin/colon and peritoneum, which respectively selected and not selected for PDGFRα‐lineage cMO.

Profiling the bulk transcriptome of inflamed skin stroma of BMT‐chimeric mice, we found several inflammation‐associated functions promoted by the Pa(lin)+ hematopoietic transplant. With the prior knowledge of the better extravasation capacity of PDGFRα‐lineage monocytes, we attempted to search for affected functions independent of extravasation. Nevertheless, regression analyses using a model separating extravasation and leukocyte lineage origin suggested PDGFRα‐lineage leukocytes had impacted the dermal environment mainly via extravasation. While a portion of extravasated monocytes proceeds to macrophage differentiation, trMF are amongst the leukocytes having extensive crosstalk with tissue stromal cells and have been reported to facilitate tissue remodelling [44, 45, 46]. Monocyte‐derived macrophages seems to assimilate much of the bona fide trMF functionality when placed into the tissue niche [17, 47]. Having analyzed with single‐cell transcriptomics the macrophage repertoire in inflamed skin, we found macrophages of different PDGFRα‐lineage origins displayed skewed functional composition. In particular, PDGFRα‐lineage dermal macrophages seem to receive stronger contribution from circulating monocytes as they contain a smaller fraction of Itgax + Tgfbr1 + Runx3 hi cells (subtype 2) resembling yolk sac‐derived Langerhans cells [48, 49, 50]. With these insights, we propose extravasated PDGFRα‐lineage monocytes differentiate into macrophages with an altered functional profile and further interact with the dermal environment.

Collectively, we document BM monocytes descended from PDGFRα+ embryonic precursors dominantly support macrophage formation in peripheries but not in BM. PDGFRα‐lineage BM F4/80+SSClo monocytes mostly occupy the Ly6C+ compartment where high integrin expression facilitates adhesion on endothelium and circulation entry. This event routes more PDGFRα‐lineage monocytes into peripheries and drives spatial and functional divergence from non‐PDGFRα‐lineage in macrophages. Our findings imply the diverse embryonic lineage origins of leukocytes might be a determinant in forming the immune functional heterogeneity in adults.

Materials and methods

Animals

Generation of PDGFRα‐lineage tracing mice (PDGFRα‐CRE;ROSAtdTomato) was previously described [34]. PDGFRα‐H2BGFP knock‐in reporter mouse line was previously described [22]. EGFP‐Tg mice were purchased from Jackson Laboratory (Tg(act‐EGFP)Y01Osb, #006567) [51]. Unless otherwise specified, 7‐ to 14‐week‐old mice of C57BL/6 background (CLEA Japan) were used for experiments. For dermal inflammation, only female mice were used to avoid sporadic inflammation related to fighting/injury. Animals were maintained in a barrier facility under special pathogen‐free conditions. All animal experiments were conducted under the guidelines of Osaka University and approved by the Institutional Animal Care and Use Committee of Osaka University Graduate School of Medicine under reference numbers 29‐044‐067 and 02‐075‐002.

Cells

Primary dermal ECs (MDMVEC) were isolated and cultured as previously described [52]. To obtain marrow cells (BMC) from femur or tibia, bones were loaded to a pierced 0.6‐mL tube on top of a 1.5‐mL Eppendorf tube. Centrifugal force of 3000g was applied for 15 s to obtain cell pellets that were resuspended in PBS containing 2% FBS (resuspension buffer, RB). Vertebra was dissected and crushed in RB. Cells were pelleted and resuspended in RB. To isolate cells from dorsal skin, dissected skin was finely minced with scissors and digested in 0.3% Collagenase A (Roche), 10 U/mL DNase‐I (Roche), 0.5 mg/mL dispase‐II, 2% FBS in RPMI with agitation at 37°C for 1 h. The digest was passed through a cell strainer of 70‐micron pore (Falcon) and resuspended in RB. To isolate cells from colon, fecal contents were first removed from the dissected colon, epithelial cells were then depleted by incubating the colon in HBSS containing 5 mM EDTA with agitation at 37°C for 20 min. The colon was minced and digested in 0.1% Collagenase D (Roche), 30 U/mL DNase‐I (Roche), 0.5 mg/mL dispase‐II, 2% FBS in RPMI with agitation at 37°C for 1 h. The digest was then passed through a 70‐micron pored cell strainer (Falcon) and resuspended in RB.

Dermal inflammation

To induce acute dermal inflammation, 1 μg of LPS or saline ctrl was administered intradermally. After overnight, skin samples were collected for flow cytometric analysis. To induce atopic dermititis, bidaily/tridaily doses of 5 nmol MC903 in ethanol were applied topically on shaved dorsal skin for six times in 2 weeks. Inflamed ellipse area was monitored by 0.25π (width) × (length). On day 14, skin samples were collected from inflamed area and prepared for flow cytometric analysis.

Peritoneal inflammation

Mice received a single intraperitoneal dose of 500 ng rm‐TNF‐α (PeproTech). After overnight, peritoneal lavage was performed by injecting 5mL PBS containing 1 mM EDTA into PC with gentle rubbing, volume was recovered afterward and prepared for flow cytometric analysis.

BM transplantation

Cells were isolated from femur/tibia of PDGFRα‐lineage tracing mice. After bead‐removal of lineage+ cells, Pa(lin)+ and Pa(lin)‐ cells were sorted by FACSAria‐III (BD Biosciences) and intravenously transplanted to irradiated EGFP‐Tg recipients (10 Gy) as previously described [5]. Six weeks were allowed for BM reconstitution before experimentation.

Three‐dimensional confocal microscopy

Freshly isolated femurs from PDGFRα‐lineage tracing mice were prepared into frozen tissue block as previously described [53] except the fixation was extended to overnight. Sixty‐micron thick cryosections were prepared on slides. After rehydration in PBS, the tissue was permeabilized in 0.5% Triton‐X‐100 in PBS for 15 min and then blocked in 3% BSA and 0.3% Triton‐X‐100 in PBS (diluent) for 1 h at room temperature. Tissues were stained with 10 μg/mL each of anti‐F4/80‐Alexa‐Fluor‐488 (BM8), anti‐PECAM‐1‐Alexa‐Fluor‐647 (390 and Mec13.3), and anti‐Sca1‐Brilliant‐Violet‐421 (D7) (all from Biolegend) in diluent at 4°C in the dark overnight. The slides were then washed and mounted for image stack acquisition by LSM880 (Zeiss). Fiji Image J was used for 3D‐stack examination of adherent monocytes on endothelium. Gamma (0.85) was applied to the tdTomato channel for simultaneous visualization of bright perivascular cells and dimmer PDGFRα‐lineage leukocytes in BM. For examination of AD skin, samples were process similarly as described except cryosections were prepared in 100 μm‐thick and stained with 10 μg/mL each of anti‐F4/80‐Alexa‐Fluor‐488 (BM8) and anti‐PECAM‐1‐Alexa‐Fluor‐647 (Mec13.3) (both from Biolegend).

EdU pulse tracing

One milligram of EdU was injected to mice via intraperitoneal route. EdU‐labeled cells were detected by click chemistry with EdU in vivo Kit (Baseclick GmbH) according to manufacturer protocol with modifications. BMC were isolated from femur/tibia and resuspended in 1 mL per limb PBS containing 2% FBS and 200 μL cells were used for each staining/reaction. Standard flow cytometric staining procedures (described in the section “Flow cytometry”) were performed. All subsequent procedures were performed at room temperature in the dark. Cells were fixed with 100 μL 4% PFA for 15 min followed by wash. Permeabilization was performed with 100 μL 0.6× permeabilization buffer (Baseclick GmbH) for 20 min. Click reaction concentrate (2×) was freshly prepared and 100 μL was directly added to cells followed by quick mixing. The reaction mixture was incubated for 30 min followed by washes with 0.6× permeabilization buffer. Samples were resuspended in 0.3× permeabilization buffer and analyzed with a flow cytometer (BD FACSCanto II, BD Biosciences).

Competitive adhesion assays

For adhesion on MDMVEC, 96‐well plate was coated with 5 μg/mL each of LN411 and LN511 (Biolamina) at 37°C for 1 h. Twenty‐thousand WT MDMVEC were seeded per well and cultured to confluence. MDMVEC was stimulated with 0 or 5 nM recombinant murine TNF‐α (Peprotech) overnight. BMC obtained from femur, tibia, or vertebra of PDGFRα‐lineage tracing mice were resuspended in DMEM buffered with 10 mM HEPES at pH7.0 (assay buffer). Fc receptors were blocked by incubating with 10 μg/mL 2.4G2 at room temperature for 10 min. MDMVEC were washed with assay buffer followed by addition of 2 × 105 BMC labelled with anti‐F4/80‐APC in 100 μL assay buffer. Cells were allowed to adhere at 37°C for 30 min. Plate was washed for trice to remove nonadherent cells, then the adherent cell complex was dissociated by accutase at 37°C. The proportion of Pa(lin)+ F4/80+ monocytes in the adherent cells was compared to input by flow cytometry. For adhesion on CAMs, 96‐well plate was coated with 5 μg/mL of ICAM‐1‐Fc or VCAM‐1‐Fc (both from R&D Systems) at 4°C overnight followed by blockade with 1% casein in PBS at room temperature for 1 h. BMC were prepared as described above. The plate was washed and 50 μL of 40 ng/mL CCL2 was added. Two hundred thousand labelled BMC in 50 μL assay buffer were allowed to adhere on the surface at 37°C for 30 min. Subsequent measurements were performed as described above. In some experiments, 1 or 10 mM MnCl2 was used instead of CCL2 as the stimulant.

Flow cytometry

Cells obtained from PDGFRα‐lineage tracing mice were analyzed for the proportion of Pa(lin)+ (tdTomato+) in target cells. Before staining, Fc receptors were blocked with 10 μg/mL 2.4G2 at room temperature for 10 min. Standard staining procedures were performed on ice to identify target cells using the following antibodies (all from Biolegend): anti‐F4/80‐APC or ‐FITC or ‐BV421 (BM8), anti‐MHCII‐PE/Cy7 (M5/114.15.2), anti‐Ly6C‐FITC or ‐PE (HK1.4), anti‐CD19‐FITC or ‐APC (6D5), anti‐CD24‐PE/Cy7 (M1/69), anti‐CD23‐PerCP/Cy5.5 (B3B4), anti‐IgM‐APC (RMM‐1), anti‐Ly6G‐FITC or ‐APC (1A8), anti‐Sca‐1‐PE/Cy7 (D7), anti‐Lineage cocktail‐FITC (145‐2C11, RB6‐8C5, RA3‐6B2, Ter‐119 and M1/70), anti‐Kit‐APC (2B8), anti‐PECAM‐1‐PerCP/Cy5.5 (390), anti‐LFA‐1‐PE/Cy7 or ‐PerCP‐Cy5.5 (H155‐78), anti‐Itgam‐BV421 (M1/70), anti‐Itga4‐PE/Cy7 (R1‐2), anti‐Itgb1‐APC/Cy7 (HMGb1‐1), anti‐Siglec‐1‐FITC (3D6.112), anti‐CXCR4‐PerCP/Cy5.5 (L276F12), anti‐CX3CR1‐Alexa‐Fluor‐488 (SA011F11) and anti‐CD64‐APC (X54‐5/7.1). For staining of colonic cells, antibody staining was accompanied by addition of mouse CD45 nanobeads (Biolegend) for magnetic enrichment of leukocytes from colonic digest. Cells, if not fixed, were stained by Sytox‐blue to exclude dead cells from analysis. In some experiments, MFI was measured to determine the surface expression level of the antigen. Flow cytometry was performed in compliance to the guidelines described in https://doi.org/10.1002/eji.201970107. Gating strategies were described in Supporting information Figures S1–5, 9, and 14.

Transcriptomic library preparation and analyses

For bulk RNA‐sequencing, RNA was extracted from cultured macrophages directly by RNeasy Mini Kit (Qiagen) or from AD skin (homogenized under low temperature, liquid N2) with ISOGEN (Nippon Gene) and purified by RNeasy Mini Kit (Qiagen). Library was prepared with 12.5 ng of total RNA using a KAPA RNA HyperPrep Kit with RiboErase (HMR) for Illumina (Kapa Biosystems). Briefly, total RNA was depleted for rRNA by hybridization with complementary DNA oligonucleotides, followed by treatment with RNase H and DNase to remove rRNA duplexed with DNA and original DNA oligonucleotides. After fragmenting, the RNA using heat and magnesium, first‐strand cDNA was synthesized using random priming, followed by a combined second strand synthesis and A‐tailing process. Libraries were amplified by PCR after the addition of adapters and finally size selected using magnetic beads. The concentration of DNA libraries was determined using the Qubit 3.0 Fluorometer (Thermo Fisher Scientific) and DNA size was analyzed with a 4200 TapeStation (Agilent Technologies). The resulting libraries were sequenced on NextSeq2000 (Illumina) in the paired‐end mode (read1: 37bp, index1: 8bp, index2: 8bp, read2: 37bp). Outputs were demultiplexed by bcl2fastq2 (Illumina). After trimming adapter sequences with TrimGalore (https://github.com/FelixKrueger/TrimGalore), reads were aligned to the mouse reference genome (mm10, obtained from the iGenomes repository, Illumina). Transcript expression was then quantified using RSEM with STAR. Sequencing counts were normalized by DESeq2 [54] and analyzed with R (version 4.0.3). Transcription activity was analyzed with the DoRothEA algorithm [55, 56]. For regression analysis, three parameters were determined for each sample: extravasation index (Extrvs), PDGFRα‐lineage (PaLin), and aggregated expression of enriched gene in a concerned enriched GOBP (aggExp). Extrvs is defined as the numeric product of donor‐derived CD45+ recruitment and donor‐derived F4/80+ recruitment, both normalized to recipient dermal stromal cells. PaLin is either +1 for Pa(lin)+ BMT or −1 for Pa(lin)– BMT. To compute aggExp of a gene set, DESeq2‐normalized expression was first divided by expression average of all samples for each gene (to balance the expression magnitudes); the computed values from all genes in the gene set were then averaged to give aggExp. Linear regression was performed for each enriched GOBP with the model: aggExp = a(Extrvs) + b(PaLin) + c(Extrvs x PaLin)+(intercept), where a, b, and c are coefficients. Coefficients with p > 0.1 were rejected.

For single‐cell RNA‐sequencing, single cells were prepared from AD skin of PDGFRα‐lineage tracing mice. Cells were blocked with 10 μg/mL 2.4G2 at room temperature for 10 min followed by staining with anti‐F4/80‐APC (BM8) and biotinylated anti‐MHCII (M5/114.15.2) (both from Biolegend). MHCII‐expressing cells were positively selected by streptavidin‐magnetic beads (Biolegend). Cells were stained by Sytox‐blue and viable F4/80+ Pa(lin)+ and Pa(lin)‐ cells were sorted by FACSAria‐III (BD Biosciences) for single‐cell RNA‐sequencing. Live cells were sorted into 384‐well plates (Eppendorf) using a BD FACSAria III instrument (Becton Dickinson; 100 μm chip) in single‐cell purity mode (1 cell/well). Single‐cell RNA‐seq libraries construction was previously described [34]. The libraries were sequenced on a NextSeq500 platform (Illumina). Fastq files were aligned to Mus musculus GRCm38 reference using STAR aligner (2.7.1a; https://github.com/alexdobin/STAR). STARsolo outputs were filtered using the numbers of reads, transcripts, and genes and the percentages of mitochondrial genes and mapped reads for each cell. Data analysis was performed by Monocle3 package [57, 58]. Single cells were clustered on Uniform Manifold Approximation and Projection, nonmacrophage clusters were removed and the macrophages were subclustered into subtypes. Subtype proportion of Pa(lin)+ and Pa(lin)‐ were computed.

Statistics

Two‐tailed Student's t‐test or Mann–Whitney U‐test was used to compare two groups. One‐sample t‐test was used to compare a control‐normalized group against an expected value. For multiple comparison, one‐way ANOVA or two‐way ANOVA was used followed by a posthoc test indicated for each experiment. Data normality was verified by Shapiro–Wilk test. Where applicable, data from multiple sample acquisition batches are shown as pooled summary. An α‐value of 0.05 was used to consider statistical significance.

Funding

This work was partly supported by AMED (Grant number 20ek0109400h0003, to KT) and JSPS KAKENHI (Grant number 19H03682, to KT).

Author Contributions

Conceptualization: YTL

Investigation: YTL, SY, ET, TK

Formal analysis: YTL, YO

Writing—original draft: YTL

Writing—review & editing: YTL, KT

Project administration: YTL, KT

Funding acquisition: KT

Conflict of interest

KT is a scientific founder and stockholder of StemRIM. SY, ET, TK, and YO are employees of StemRIM. All other authors declare they have no conflict of interest.

Peer review

The peer review history for this article is available at https://publons.com/publon/10.1002/eji.202149479

atopic dermatitis

Bone marrow cells

BM transplantation

cell adhesion molecules

common monocyte progenitor

endothelial cell

Gene ontology

GO biological processes

hematopoietic stem and progenitor cells

Lineage‐ Sca1+ Kit+

macrophage and DC progenitor

peripheral blood

peritoneal cavity

tissue‐resident macrophages

transcription factor

Supporting information

Click here for additional data file.