A Novel Subset of Anti-Inflammatory CD138⁺ Macrophages Is Deficient in Mice with Experimental Lupus Free

This article is featured in In This Issue, p.1211

Macrophages (Mϕ) play a key role in the noninflammatory disposal of apoptotic cells (1). Monocyte-derived Mϕ from systemic lupus erythematosus (SLE) patients are poorly phagocytic (2) and patients accumulate apoptotic cells in their tissues (3–6). Dead cells also accumulate in tissues of mice with pristane-induced lupus (6), but not in mice treated with mineral oil (MO), an inflammatory hydrocarbon that does not cause lupus. Impaired phagocytosis of apoptotic cells promotes murine lupus (7–9). Although phagocytosis is usually noninflammatory (8, 9), impaired phagocytosis of dead cells in lupus facilitates endosomal recognition of self-nucleic acids by TLR7 and TLR9, resulting in proinflammatory cytokine production (10). The outcome of phagocytosis (pro- versus anti-inflammatory) depends on the release of damage-associated molecular patterns by dying cells, whether the cells are apoptotic or necrotic, the type of phagocyte, receptors mediating uptake, and factors regulating the sorting of apoptotic cells after phagocytosis or the coupling of phagocytosis to anti-inflammatory pathways (11–14). By overwhelming normal clearance mechanisms, an increased rate of cell death also may promote lupus (15–19).

We show impaired clearance of dead cells by lupus bone marrow (BM) Mϕ and report a novel subset of peritoneal syndecan-1 (CD138)⁺ Mϕ with an anti-inflammatory phenotype that efficiently takes up apoptotic cells in the peritoneum. This subset is deficient in mice with pristane-induced lupus, resulting in impaired apoptotic cell clearance and inflammation.

BM core biopsies were identified from the University of Florida (UF) Department of Pathology archives. SLE was classified using American College of Rheumatology criteria (20, 21). Biopsies from adults with acute myelogenous leukemia (AML) undergoing myeloablation with cytarabine plus daunorubicin 14-d earlier and children with B cell acute lymphocytic leukemia (B-ALL) treated with vincristine, prednisone, anthracycline, plus cyclophosphamide and/or l-asparaginase 8-d earlier were de-identified and examined by H&E staining and immunohistochemistry (IHC). The patients were not treated with radiation and did not receive cytokines or growth factors in the week before BM biopsy. Biopsies in which marrow cellularity dropped from 100% to <5% following myeloablation were selected for further study (n = 4). BM biopsies from patients undergoing myeloablation were compared with biopsies from SLE patients (n = 6) and controls undergoing BM biopsy for staging of lymphoma who had no evidence of BM involvement (n = 6). The UF Institutional Review Board approved these studies.

BM core biopsies were fixed in 10% neutral buffered formalin and decalcified (6). Four-micrometer sections were deparaffinized and underwent heat-induced epitope retrieval before staining with anti–cleaved-caspase-3 (Cell Signaling), anti–TNF-α (Abcam), and anti-CD68 (Dako) Abs followed by peroxidase- or alkaline phosphatase–conjugated goat secondary Abs (6). Reaction product was visualized using ultraView DAB (brown) or Alkaline Phosphatase Red Kits (Ventana). Slides were counterstained with hematoxylin. Numbers of activated caspase-3⁺ cells (red) that did not colocalize with MΦ (brown) were determined as the mean number of red⁺brown⁻ cells per 100× field (4 fields per patient).

Mice were maintained under specific pathogen-free conditions at the UF Animal Facility. C57BL/6 (B6) mice (Jackson Laboratory) received 0.5 ml of pristane (Sigma), MO (C.B. Fleet), 100 ng LPS from Salmonella enterica serotype Minnesota (Sigma), or PBS i.p. At indicated times, peritoneal exudate cells (PEC) were collected by lavage. Cells were analyzed within 1 h. Bronchoalveolar lavage (BAL) was performed after euthanizing the mice. A small incision was cut in the trachea and 1-ml PBS was injected using a 20-gauge plastic feeding tube (Instech Laboratories). Lung washings were analyzed within 1 h. Animal studies were approved by the UF Institutional Animal Care and Use Committee.

Quantitative PCR (Q-PCR) was performed as described (6) using RNA extracted from 10⁶ peritoneal cells (RNA Isolation Kit; Qiagen). cDNA was synthesized using the Superscript II First-Strand Synthesis Kit (Invitrogen) and SYBR Green Q-PCR analysis was performed using an Opticon II thermo cycler (Bio-Rad). Primer sequences were: Mertk forward 5′-GAGACCTCCACACCTTCCTG-3′, reverse 5′-GAGCTGCCAAATCCCTATGA-3′; Sra1 forward 5′-CAACATCACCAACGACCTCA-3′, reverse 5′-TGTCTCCCTTTTCACCTTGG-3′; Marco forward 5′-AGCCGATTTTGACCAAGCTA-3′, reverse, 5′-GTGAGCAGGATCAGGTGGAT-3′; Srb1 forward 5′-CCGAGAGTCTGGCATTCAG-3′, reverse 5′-TGGGTTAGGGTTCAGACCAA-3′; Gata6 forward 5′-TACAAGAACACCAACACAGTCC-3′, reverse 5′-GGCGTCAAGAGTGTTACAGATAC-3′; Pparg forward 5′-CAAGGTGCTCCAGAAGATGA-3′, reverse 5′-GTGAAGGCTCATGTCTGTCT-3′; Runx3 forward 5′-CGTGTAACACCAAGCACACC-3′, reverse AAGGGGTTCAGGTCTGAGGA-3′; Car4 forward ACCTCTGACCTCAGCCTTTA-3′, reverse CCACAGCCAGTTCCTCATATT-3′; Alox15 forward 5′-GCTTTCCATAGTCTGCTGTAGT-3′, reverse 5′-TTTCCCAAGCATGGCTATTATTTAC-3′; β-actin forward 5′-TGGAATCCTGTGGCATCCTGAAAC-3′, reverse 5′-TAAAACGCAGCTCAGTAACAGTCCG-3′.

Flow cytometry was performed as described (6). Cells were incubated with anti-mouse CD16/32 (Fc Block; BD Biosciences) before staining with primary Ab or isotype controls. A total of 10,000–50,000 events per sample were acquired using an LSRII Flow Cytometer (BD Biosciences) and analyzed with FlowJo flow cytometry software (Tree Star). The following Abs were used: CD11b-Brilliant Violet 421, Tim4-PE, CD138-allophycocyanin, CD138-PE F4/80- Pacific Blue, Ly6G-allophycocyanin-Cy7, Ly6C-allophycocyanin-Cy7, CD80-PerCp-Cy5, CD40-PerCP-Cy5, CD206-PerCP-Cy5, CCR2-FITC, CX3CR1-FITC, CD169-allophycocyanin, CCR5-allophycocyanin, CD36-PE, CD36-AlexaFluor 488, and IL-10R–PE (BioLegend); Marco-FITC (Bio-Rad); Ly6C-FITC, CD86-FITC, CD11c-FITC, I-A/I-E–PE, and CD93-PE (BD Bioscience); CREB-PE and Phospho-CREB-FITC (Cell Signaling); and CD11b-Pacific Blue, CD45-FITC, TLR4-PE, and CD45-FITC (eBioscience). Buffers used in intracellular staining were from eBioscience. For cell sorting, PEC from untreated B6 mice were stained with anti-CD11b, -Tim4, and -CD138 Abs. CD11b⁺Tim4⁺ and CD11b⁺CD138⁺ cells were gated and sorted (FACSaria Cell Sorter). Forty-thousand cells per subset were collected and lysed immediately for RNA extraction.

Mouse thymus lymphoma cells (BW5147) were induced into apoptosis by heat shock (45°C, 10 min) and cultured for 4 h (22). The percentage of apoptotic (annexin-V⁺) cells was routinely >80% by this technique. After inducing apoptosis, cells were washed twice and labeled with pHrodo Red (Life Technologies) (23). Uptake of labeled apoptotic cells was assayed in vitro by adding 10⁶ apoptotic cells in AIM-V medium to each well of a 12-well plate containing 10⁶ adherent peritoneal cells. Plates were incubated for 2.5 h at 37°C, washed three times with ice-cold PBS, and stained with Pacific Blue–conjugated rat anti-mouse CD11b Ab (30 min, 4°C). Cells were washed again, detached with 1-ml ice-cold PBS using a cell scraper, and fixed with 1% paraformaldehyde before flow cytometry.

For in vivo assays, 2 × 10⁷ pHrodo Red–labeled apoptotic cells were injected i.p. into pristane- or MO-treated mice (1 wk after treatment). After 1.5 h, PEC were collected by lavage; stained with fluorescently labeled anti-CD11b, anti-Ly6C, anti-CD138, and anti-Ly6G Abs; washed; and fixed as above.

To assess the role of Marco in apoptotic cell uptake, IgG anti-Marco neutralizing Ab (ED-31; AbD Serotec) or isotype control were injected i.p. into MO-treated mice (100 μg/mouse). Thirty minutes later, pHrodo Red–labeled apoptotic cells were injected and uptake was determined 1.5 h later as above. The class A scavenger receptor inhibitor polyinosinic acid (Poly-I) (Sigma) was used to further confirm the role of Marco. Poly-I (200 μg/mouse) or PBS was injected i.p and 30 min later, pHrodo Red–labeled apoptotic BW5147 cells were injected. Uptake was measured by flow cytometry 1.5 h later.

In vitro phagocytosis of unopsonized polystyrene beads (3.2-μm diameter) was measured by flow cytometry. PEC were obtained 1 wk after MO treatment and cultured for 1 h with PE-labeled polystyrene beads at a 10:1 bead/cell ratio (DakoCytomation). Cells then were stained with anti-CD11b, -Ly6C, -Ly6G, and -CD138 Abs. The percentage of CD11b⁺Ly6G⁻Ly6C⁻CD138⁺ and CD11b⁺Ly6G⁻Ly6C⁻CD138⁻ cells that had taken up PE-labeled beads was determined by flow cytometry.

Peritoneal cells from untreated mice were incubated with PBS or LPS (100 ng/ml) for 30 min. Total CREB and p-CREB staining was determined by flow cytometry in the CD11b⁺CD138⁺Tim4⁻, CD11b⁺CD138⁻Tim4⁺, and CD11b⁺CD138⁻Ly6C^hi subsets. In some experiments, mice were treated with the cell-permeable adenylate cyclase inhibitor SQ22536 (24). Mice received MO (0.5 ml i.p.) on day 1, and SQ22536 (250 μg/d i.p) (25) for 9 d starting on day 1. Peritoneal cells were analyzed by flow cytometry at day 9.

Statistical analyses were performed using Prism 6.0 software (GraphPad Software). Differences between groups were analyzed by unpaired Student t test (normally distributed data) or Mann–Whitney U test (non-normally distributed data). Data are expressed as mean ± SD for normally distributed data sets. Normality was determined by the D’Agostino and Pearson omnibus test. All tests were two sided and a p value <0.05 was considered significant.

Dead cells accumulate in BM and lung tissue of mice with pristane-induced lupus and SLE patients, and TNF-α is produced locally (6, 26). To see if these abnormalities are associated with a phagocytosis defect, we compared BM from SLE patients and controls.

SLE BM contained cells staining with Abs against activated caspase-3, an apoptosis marker (Fig. 1A, middle). In contrast, caspase-3⁺ cells were largely absent in BM from a non-SLE control (Fig. 1A, left). Wright staining also showed numerous dead cells in patients’ BM aspirates (Fig. 1A, right).

To assess the clearance of dead cells, we studied patients undergoing myeloablative therapy for AML and B-ALL, which causes massive death of BM cells (27). BM from a child with B-ALL exhibited 100% cellularity prior to myeloablative therapy, but 8-d later was hypocellular (<5% cellularity) with erythroid regeneration (Fig. 1B). Large phagocytic cells, probably resident BM Mϕ, were visualized in the hypocellular marrow by H&E staining and IHC showed activated caspase-3–stained material within endosomes/phagosomes (Fig. 1B). Similar results were obtained using day-14 BM from AML patients (data not shown). Caspase-3⁺ material was restricted to endosomes or phagosomes of CD68⁺ Mϕ (Fig. 1C, top left), whereas in SLE patients activated caspase-3 was found in intact cells located outside of CD68⁺ BM Mϕ (Fig. 1C, right). Quantification of the number of caspase-3⁺ cells outside of CD68⁺ Mϕ confirmed that dead (caspase-3⁺) cells were removed inefficiently in SLE BM versus BM from non-SLE patients undergoing myeloablative therapy (p < 0.01, Student t test) (Fig. 1D). There was no significant difference in caspase-3⁺ cells in BM from myeloablated patients versus untreated controls. Human lupus BM also stained prominently for TNF-α (Fig. 1E, bottom right) (6), whereas TNF-α was undetectable in BM from B-ALL 8 d after myeloablation (Fig. 1E, bottom left). Thus, nonlupus BM has an enormous capacity to clear dead cells but clearance is impaired and there is TNF-α production in BM from SLE patients (Fig. 1) and pristane-treated mice (6). To explore the mechanism, we assessed the clearance of dead cells in pristane-induced lupus.

In vitro phagocytosis assays revealed that CD11b⁺ PEC from MO-treated mice took up 2.5-fold more apoptotic cells than CD11b⁺ PEC from pristane-treated mice (p < 0.01, Student t test) (Fig. 2A). The mean fluorescence intensity (MFI) of pHrodo Red in CD11b⁺ cells also was lower in pristane- versus MO-treated mice. Because in vitro culture alters Mϕ function, in vivo phagocytosis was measured by injecting pHrodo Red–labeled apoptotic cells into the peritoneum followed 1.5 h later by flow cytometry for internalized target cells. Consistent with the in vitro data, a higher proportion of CD11b⁺ cells from MO-treated mice took up apoptotic target cells and the pHrodo Red MFI was higher in comparison with pristane-treated mice (Fig. 2B). We conclude that the uptake of dead cells is impaired in pristane-induced lupus, as in SLE patients.

In nontreated mice, the peritoneum contains a population of large, strongly phagocytic CD11b⁺Tim4⁺-resident Mϕ (28). Because these cells disappear in certain forms of inflammation (29–31), we examined whether differences in their numbers might explain the impaired phagocytosis of apoptotic cells in pristane- versus MO-treated mice. However, CD11b⁺Tim4⁺-resident Mϕ disappeared 7 d after i.p. injection of either pristane or MO (Fig. 2C). Thus, although Tim4⁺-resident Mϕ are highly phagocytic, a differential effect of pristane and MO on this subset was unlikely to explain impaired phagocytosis of apoptotic cells in pristane-treated mice.

In addition to resident Mϕ, normal peritoneum contains BM-derived Mϕ (32, 33). We found an unusual subset of CD11b⁺CD138⁺ cells in the peritoneum of mice treated with pristane or MO (Fig. 3). These cells were more numerous in MO- versus pristane- or PBS-treated mice (Fig. 3A), and they expanded after MO (and to a lesser degree pristane) treatment (Fig. 3B). In contrast, CD11b⁺Ly6C^hi (inflammatory) monocytes rapidly increased in both pristane- and MO- treated mice, but decreased substantially over 14 d in MO-, but not pristane-, treated mice. At 1 wk, neutrophils increased to a similar degree in pristane- and MO-treated mice (data not shown). Thus, MO-treated mice exhibited an increasing predominance of CD11b⁺CD138⁺ cells, whereas in pristane-treated mice CD11b⁺Ly6C^hi cells predominated.

Although CD138 is usually thought of as a plasma cell marker, CD11b⁺CD138⁺ cells took up pHrodo Red–labeled apoptotic cells more actively in vivo than Tim4⁺ Mϕ from the same mouse (Fig. 3C). In addition, in vivo uptake of apoptotic targets by CD11b⁺Tim4⁻CD138⁺ cells was considerably higher than that of CD11b⁺Tim4⁻CD138⁻ cells (p < 0.001, Student t test) (Fig. 3D). Flow cytometry of PEC from MO-treated mice revealed four subsets of CD11b⁺ cells, which are best visualized in BALB/c mice but also seen in B6 mice (Fig. 3E): CD11b⁺Ly6C^hi (region [R]1), CD11b^loLy6C^neg (R2), CD11b^hiLy6C^lo/− (R3), and CD11b⁺Ly6C^int neutrophils, which also were Ly6G⁺ (data not shown). The main subset expressing CD138 was R3, which also expressed the Mϕ-restricted scavenger receptor Marco (Fig. 3E). CD138⁺Marco⁺ cells (R3) were considerably more phagocytic for apoptotic cells than other subsets (Fig. 3F).

In addition, we incubated peritoneal cells from 1-wk, MO-treated mice for 1 h with PE-labeled polystyrene beads (3.2 μm, bead/cell ratio of 10:1) and measured uptake by flow cytometry. Unopsonized polystyrene particles with a diameter >0.5 μm are taken up via Marco-mediated phagocytosis, whereas smaller particles are endocytosed (34). CD11b⁺CD138⁺ cells took up the beads more actively than CD11b⁺CD138⁻ cells (13.2% of cells versus 5.4%, p < 0.001, Student t test) (data not shown), suggesting that the high phagocytic activity of CD11b⁺CD138⁺ is not limited to uptake of apoptotic cells. However, the uptake of polystyrene particles was less efficient than uptake of apoptotic cells (Figs. 2, 3).

Levels of Marco mRNA, encoding a class A scavenger receptor involved in uptake of apoptotic cells (35), was increased in PEC from MO- versus pristane-treated mice (p < 0.005, Student t test). Expression of the class B scavenger receptor Srb1 also was increased (p < 0.02). In contrast, levels of Mertk and Sra1 mRNA were comparable (Fig. 3G). The role of Marco in clearing apoptotic cells was examined by treating mice with MO followed by i.p. injection of anti-Marco neutralizing Abs or an isotype control before measuring in vivo phagocytosis (Fig. 3H, left). Blocking Marco inhibited apoptotic cell uptake, suggesting that Marco is involved in internalization. Peritoneal injection of the class A scavenger receptor antagonist poly(I) also inhibited uptake of labeled targets by CD11b⁺ cells (Fig. 3H, right). Taken together, the data suggest that impaired phagocytosis of apoptotic cells in pristane- versus MO-treated mice may reflect the relative numbers of highly phagocytic CD11b⁺CD138⁺ cells, and expression of Marco and possibly other scavenger receptors by these cells promotes the clearance of dead cells. In contrast to the disappearance of Tim4⁺-resident Mϕ, CD11b⁺CD138⁺ cells expanded during inflammation. Although CD11b⁺CD138⁺ cells have been reported (36), relatively little is known about them.

The peritoneum contains two distinct subsets of Mϕ: resident large peritoneal Mϕ (LPM) and BM-derived small peritoneal Mϕ (SPM). The relationship of the Tim4⁺ Mϕ and CD11b⁺CD138⁺ cells in the peritoneum of untreated mice to LPM and SPM was examined by gating on CD11b⁺ cells and comparing surface-staining characteristics of the Tim4⁺ versus CD11b⁺CD138⁺ cells (Fig. 4A). Tim4⁺ Mϕ expressed more CX3CR1 than CD11b⁺CD138⁺ cells and expressed both CD169 (Siglec-1) and CD93, which were nearly undetectable on CD11b⁺CD138⁺ cells (Fig. 4A). In contrast, CD11b⁺CD138⁺ cells expressed more CCR2, CCR5, and MHC class II (MHCII) than Tim4⁺ Mϕ and also expressed CD11c, which was absent on Tim4⁺ Mϕ. The two subsets exhibited similar forward scatter, but side scatter was lower in the CD11b⁺CD138⁺ subset (Fig. 4B). The surface-staining and forward-/side-scatter properties of CD11b⁺CD138⁺ cells are similar to those reported for SPM (33).

Peritoneal cells from untreated B6 mice were stained with Abs to F4/80, CD11b, Tim4, and CD138 (Fig. 4C). About 60% of peritoneal CD11b⁺ cells were Tim4⁺CD138⁻-resident Mϕ, whereas ∼10% were CD138⁺Tim4⁻ and ∼30% were Tim4⁻CD138⁻. Consistent with the high expression of F4/80 on resident peritoneal Mϕ (32), the Tim4⁺CD138⁻ cells were F4/80^hi. In contrast, the CD138⁺Tim4⁻ subset was F4/80^int (Fig. 4C). Gene expression differed in flow-sorted peritoneal CD11b⁺F4/80^hiCD138⁻Tim4⁺-resident Mϕ versus the CD11b⁺F4/80^intCD138⁺Tim4⁻ subset (CD138⁺ Mϕ) from untreated mice (Fig. 4D). In comparison with Tim4⁺ Mϕ, CD138⁺ Mϕ expressed little Gata6 (specific for resident peritoneal Mϕ), Pparg (inducible in peritoneal Mϕ), Runx3, and Car4 (inducible in peritoneal Mϕ). Alox15 encodes 12/15-lipoxygenase, an enzyme involved in noninflammatory clearance of apoptotic cells by resident peritoneal Mϕ (13, 37) and resolution-phase Mϕ (38). It is upregulated by the uptake of apoptotic cells (39). However, although high levels were found in the Tim4⁺ Mϕ subset, CD138⁺ Mϕ expressed little Alox15 mRNA (Fig. 4D). In contrast, CD138⁺ Mϕ expressed higher levels of Marco mRNA than Tim4⁺ Mϕ.

Although present in the peritoneum, CD138⁺ Mϕ were also found in spleen from MO-treated and (in low numbers) pristane-treated mice (Fig. 5A). They were absent in spleen from PBS-treated mice (Fig. 5A). Spleen contains CD138⁺ plasma cells, but the CD138⁺CD11b⁺ population expressed CD36, F4/80, CD80, IL-10R, and Ly6C (Fig. 5B); verifying that these were CD138⁺ Mϕ. CD138⁺ Mϕ were present at low levels in BM from PBS-treated mice and did not change after pristane or MO treatment (Fig. 5C). Analysis of cells collected by BAL from MO-treated mice revealed two populations of F4/80⁺CD11b⁺ lung Mϕ: large CD138^loLy6C^lo Mϕ (R2) and smaller CD138^hiLy6C^hi Mϕ (R1) (Fig. 5D). Both expressed IL-10R, but expression was higher in R1. The large (R2) cells are likely to be alveolar Mϕ, whereas the small cells are similar to peritoneal CD138⁺ Mϕ (but expressed higher levels of Ly6C than peritoneal CD138⁺ Mϕ). BAL from pristane- or PBS-treated mice contained few CD138⁺ Mϕ (data not shown). Thus, CD138⁺ Mϕ are not restricted to the peritoneum and may develop from CD138⁻ BM precursors upon migration to sites of inflammation.

To see if the induction of CD138⁺ Mϕ is relevant to infection, mice received LPS (100 ng i.p.) and PEC were analyzed by flow cytometry. At day 7 the peritoneum of LPS-treated mice contained variable numbers of Ly6G⁺ neutrophils and 20–40% CD138⁺ Mϕ, but few (2–5%) Ly6C^hi inflammatory monocytes (Fig. 6A, 6B). LPS depleted Tim⁺CD138⁻-resident peritoneal Mϕ within 1 d (Fig. 6C). At day 3, small numbers of Tim4⁺ cells were detected in MO-treated mice and larger numbers in LPS-treated mice. Unexpectedly, at day 7, many Tim4⁺ cells in LPS-treated, but not MO-treated mice, were CD138⁺. At 14–28 d, MO-treated mice also had peritoneal Tim4⁺CD138⁺ cells (Fig. 6C). Tim4⁻CD138⁻ cells were very poorly phagocytic for apoptotic cells; whereas Tim4⁺CD138⁻, Tim4⁺CD138⁺, and Tim4⁻CD138⁺ cells could all take up apoptotic cells (data not shown). However, there was not a significant difference in the ability of the three latter subsets to take up apoptotic cells.

CD138 is a transmembrane proteoglycan encoded by Sdc1 (40). Although expressed mainly on epithelial, mesenchymal, and plasma cells, it also is found on monocytes (41). Peritoneal methylated BSA–induced F4/80^int Mϕ express CD138 and may promote the resolution of inflammation (42–44). Consistent with that possibility, IL-10R surface staining was higher on CD138⁺ Mϕ than on Ly6C^hi monocytes or CD138⁻ Mϕ in both pristane- and MO-treated mice (Fig. 7A, 7B), and CD138⁺ Mϕ exhibited lower intracellular TNF-α staining than Ly6C^hi monocytes (Fig. 7C). CD138⁺ Mϕ from untreated mice also expressed high levels of the M2 Mϕ marker CD206 (mannose receptor) and CD138⁻Tim4⁺-resident Mϕ expressed lower levels (Fig. 7D). Thus, CD138⁺ Mϕ have a phenotype (low TNF-α production, high IL-10R, and high CD206) suggestive of anti-inflammatory/resolution-phase Mϕ.

Like CD138⁺ Mϕ, resolution-phase Mϕ express M2 markers, respond poorly to TLR ligands, and are highly phagocytic (39, 45). Because resolution-phase Mϕ markedly downregulate CD11b following the uptake of apoptotic cells in vitro (39), we examined CD11b expression on CD138⁺ Mϕ. The expression level of CD11b was lower on CD138⁺ Mϕ from MO- versus pristane- treated mice (Fig. 7E, left). Following 20-h coculture of CD138⁺ Mϕ from MO-treated mice with apoptotic BW5147 cells (apoptotic targets/Mϕ ratio 5:1), CD11b fluorescence intensity substantially decreased (Fig. 7E, right). Thus, like resolution-phase Mϕ, CD138⁺ Mϕ also downregulate CD11b after taking up apoptotic cells. These data suggest that the improved phagocytic capacity in CD138⁺ cells from MO-treated mice drives increased reprogramming to the CD11b^low phenotype.

The phenotype of resolution-phase Mϕ is controlled by cAMP (45) and the cAMP/protein kinase A (PKA)-activated transcription factor CREB induces an anti-inflammatory gene expression program in Mϕ (46). As CD138 (Sdc1) is CREB regulated (41), we investigated CREB activation in CD138⁺ Mϕ. Total CREB levels were similar in CD11b⁺CD138⁺Tim4⁻ and CD11b⁺CD138⁻Tim4⁺ Mϕ from untreated mice but activated CREB (p-CREB) was increased in CD11b⁺CD138⁺Tim4⁻ Mϕ (Fig. 8A). CD11b⁺CD138⁺Tim4⁻ Mϕ also expressed higher levels of p-CREB after LPS treatment (30 min in vitro). Similar to CD138⁺ Mϕ, CD11b⁺CD138⁻Tim4⁺-resident Mϕ exhibited higher levels of p-CREB after LPS treatment. (Fig. 8B, left). Compared with Tim4⁺-resident Mϕ, CD138⁺ Mϕ from untreated mice expressed higher levels of p-CREB (Fig. 8B, right). Both phosphorylated and total CREB were higher in CD138⁺ Mϕ versus Ly6C^hi monocytes (Fig. 8C). CD11b⁺CD138⁺ peritoneal Mϕ from mice treated daily with the adenylate cyclase inhibitor SQ22536 had higher surface Ly6C staining than controls and produced more TNF-α (intracellular staining), suggesting that treatment promoted a proinflammatory phenotype (Fig. 8D). SQ22536 treatment had no significant effect on either the fluorescence intensity of CD138 staining or the percentage of CD138⁺ Mϕ (data not shown).

Phagocytosis of apoptotic cells by SLE patients’ monocytes is impaired in vitro (2) and dead cells accumulate in their tissues (3–5) and also in tissues of mice with pristane-induced lupus (6, 47). The present study suggests that the accumulation of dead cells occurs in the setting of impaired clearance by Mϕ. Reduced clearance of apoptotic cells in lupus mice was associated with low numbers of a novel subset of CD138⁺ Mϕ that was highly phagocytic for apoptotic cells and had an anti-inflammatory phenotype. Phagocytosis was mediated partly by the scavenger receptor Marco. CREB, which regulates CD138 expression and induces an anti-inflammatory transcriptional program, was activated in these CD138⁺ Mϕ (41, 46, 48).

Most tissues contain self-renewing resident Mϕ derived from embryonic precursors that seed the tissues before birth (49–51). During inflammation, resident Mϕ are transiently complemented by short-lived recruited monocytes that differentiate in situ into Mϕ (49). The accumulation of dead cells in patients’ tissues (3–6) suggests that phagocytosis is abnormal, cell death is increased, or both. Cell death is increased in SLE patients (15–17) and mice with pristane-induced lupus (18, 19). But dead cells were efficiently cleared by BM Mϕ in nonlupus patients undergoing myeloablative therapy without triggering TNF-α production (Fig. 1), consistent with other evidence that apoptotic cells are cleared efficiently without inducing inflammation (52). In contrast, dead cells accumulate in lupus BM and induce local TNF-α production (Fig. 1) (6). In pristane-induced lupus, phagocytosis of apoptotic cells was impaired in vitro and in vivo (Fig. 2). This was not seen in mice treated with MO, an inflammatory but nonlupus-inducing hydrocarbon oil. Defective apoptotic cell removal may contribute to the pathogenesis of lupus by activating innate immunity (7).

CD11b⁺Tim4⁺-resident peritoneal Mϕ, which actively take up apoptotic cells, were reduced by >90% shortly after pristane or MO treatment, consistent with prior studies showing that peritoneal inflammation rapidly depletes resident peritoneal Mϕ (29, 53). This “Mϕ disappearance reaction” (31) is caused by relocalization of CD11b⁺Tim4⁺ (Gata6⁺)-resident peritoneal Mϕ to omental milky spots (30). CD138⁺ Mϕ were more active than CD11b⁺Tim4⁺ Mϕ at taking up apoptotic cells and were not depleted by peritoneal inflammation, but were affected differentially by pristane and MO. In MO-treated mice, CD138⁺ Mϕ expanded and Ly6C^hi monocytes decreased, whereas high levels of Ly6C^hi monocytes and lower levels of CD138⁺ Mϕ were maintained in pristane-treated mice (Fig. 3). CD138⁺ Mϕ were the most active peritoneal myeloid subset at taking up apoptotic cells, and their relative deficiency may explain the decreased clearance of apoptotic cells in pristane- versus MO-treated mice. CD138⁺ Mϕ were not unique to pristane- or MO-induced inflammation, as they appeared in the peritoneum 7 d after LPS treatment (Fig. 6). Thus, CD138⁺ Mϕ may play a physiological role in resolving infectious as well as sterile inflammation, possibly by promoting clearance of dead cells.

Impaired phagocytosis of apoptotic cells promotes murine lupus (7). Two phagocytic processes remove dead cells: opsonin-dependent (e.g., Fc receptors, complement receptors, and TAM receptors) and opsonin-independent (including phosphatidylserine receptors, such as Tim4) processes (9). Opsonins such as MFG-E8, Gas6, and protein S are involved in the recognition of apoptotic cells by integrins and the TAM receptors (54–56). This plays a major role in clearing apoptotic cells. In contrast, Marco and other scavenger receptors recognize oxidized low-density lipoprotein on the surface of apoptotic cells without a requirement for opsonization, promoting their clearance and inhibiting cytokine/chemokine responses (35, 57, 58). CD138⁺ Mϕ expressed the class A scavenger receptor Marco and class B scavenger receptors SR-B1 and CD36. Treatment with anti-Marco neutralizing Abs or the class A scavenger receptor antagonist Poly-I reduced phagocytosis of apoptotic cells by ∼1/3 (Fig. 3), suggesting that clearance involves Marco. CD138⁺ Mϕ also took up polystyrene particles more actively than CD138⁻ Mϕ. Consistent with the role of Marco in clearing apoptotic cells, phagocytosis of unopsonized particles also is Marco dependent (34).

Marco promotes noninflammatory phagocytosis of apoptotic cells and is expressed constitutively on marginal zone Mϕ and some tissue Mϕ. It is induced on other Mϕ subsets via TLR signaling (9, 51). A Marco defect contributes to lupus in BXSB male mice (59), and autoantibodies against Marco impair apoptotic cell uptake in murine and human lupus (57, 60). Moreover, anti–DNA autoantibody production is increased in Marco-deficient mice following injection of apoptotic cells (57). Our data support the idea that Marco deficiency promotes lupus. Expression of SR-B1 also was low in pristane-treated mice (Fig. 3). Because human SR-B1 recognizes apoptotic cells (61), it will be of interest to see if it plays a similar role in mice.

Normal peritoneum contains a major population of self-renewing resident LPM established during fetal life and a minor (∼10%) population of BM-derived SPM (33). Gata6, a master transcriptional regulator of resident peritoneal Mϕ (29, 30), was expressed at high levels by Tim4⁺CD138⁻ LPM, as expected (Fig. 4D). The near absence of Gata6 expression in Tim4⁻CD138⁺ Mϕ and intermediate F4/80 staining (Fig. 4C) are consistent with BM-derived SPM (32, 33). Their surface phenotype further supports the close relationship of CD138⁺ Mϕ to SPM (33). Both express high levels of MHCII and low or intermediate levels of F4/80, but not CD93 (AA4.1) or Ly6C (Fig. 4). In contrast to LPM, CD138⁺ Mϕ (Figs. 2, 3) and SPM (33) do not disappear from the peritoneum during inflammation. However, the kinetics of appearance/disappearance of F4/80^hiCD11b^hi (LPM-like) and F4/80^intCD11b^int (SPM-like) cells may vary depending on the model of peritoneal inflammation used (62).

In contrast to previous reports, our studies suggest that CD138⁺ Mϕ or SPM are not restricted to the peritoneum, as they were detected in the spleen and lungs of pristane- and MO-treated mice (Fig. 5). They did not change in the BM after treatment, suggesting that CD138⁺ Mϕ arise peripherally by differentiation of CD138⁻ BM precursors. Ly6C^hi monocytes can give rise to SPM during inflammation, suggesting that Ly6C⁺ cells expressing low levels of CD138 (Fig. 3E) and possibly the CD138⁺Ly6C⁺ lung Mϕ (Fig. 5D) are an intermediate stage in the differentiation of CD138⁺Ly6C⁻ Mϕ from CD138⁻Ly6C^hi precursors.

A population of CD138⁺Tim4⁺ Mϕ was found in LPS-treated mice (Fig. 6C) and similar cells appeared later on in MO-treated mice. These cells resemble a subset of alternatively activated resident Mϕ (63). However, further studies are needed to confirm that they are not BM derived.

Although an oversimplification (64), BM-derived Mϕ have been classified into two subtypes: M1 (classically activated, proinflammatory) and M2 (alternatively activated, associated with resolution). M1/M2 polarization is influenced by transcription factors (64, 65) and epigenetic regulation (66). Mouse M1 Mϕ express iNOS (Nos2) and high levels of Ly6C, MHCII, and CCR2, and produce TNF-α, IL-1β, and IL-12; whereas M2 Mϕ express arginase 1 (Arg1), scavenger receptors, CX3CR1, and low levels of Ly6C, and produce TGF-β and IL-10 (66). CD138⁺ Mϕ exhibited some features of both. Consistent with an M1 phenotype, they expressed high levels of MHCII and CCR2 and low levels of CX3CR1 (Fig. 4). But they were Ly6C⁻ and expressed CD206 (Fig. 7), an M2c Mϕ marker expressed on BM-derived, but not resident, M2-like Mϕ (63, 67, 68). Moreover, like other proresolving Mϕ (39), CD138⁺ Mϕ downregulated CD11b, a component of complement receptor 3, following coculture with apoptotic cells (Fig. 7E). In blood vessels, CD138⁺ Mϕ are anti-inflammatory (36). A CREB and cAMP/PKA-regulated transcriptional program controls Sdc1 (CD138), Klf4, Arg1, Il10, and other genes involved in M2 polarization (41, 46, 48). The increased levels of p-CREB in CD138⁺ Mϕ and high levels of IL-10R expression are consistent with an M2-like phenotype (Fig. 7). We speculate that the pattern of chemokine receptor expression (CCR2^hi CX₃CR1^lo) might allow anti-inflammatory CD138⁺ Mϕ to migrate to the same sites as proinflammatory Ly6C^hi (CCR2^hi) monocytes, promoting resolution of inflammation. Thus, CD138⁺ Mϕ exhibit several features of resolution-phase Mϕ, including an M2-like surface phenotype with some M1 markers, high phagocytic activity, and downregulation of CD11b following the uptake of apoptotic cells (39, 45).

Finally, our studies suggest that the cAMP/PKA-regulated transcription factor CREB may help maintain the M2-like or proresolving phenotype of CD138⁺ Mϕ. The increased level of p-CREB in CD138⁺ Mϕ supports that possibility (Fig. 8). CREB activation promotes M2 polarization (46, 69) and the phenotype of resolution-phase Mϕ is controlled by cAMP (45). Although CD138 expression is regulated by cAMP/PKA and CREB (41), we did not see lower CD138 surface staining in mice treated with the adenylate cyclase inhibitor SQ22536 (data not shown). However, Ly6C staining and TNF-α were increased, consistent with a shift toward M1-like polarization (Fig. 8D). It remains unclear whether CD138 is a component of a transcriptional program that dampens inflammatory responses (36), or if it is only a marker for a subset of anti-inflammatory/resolution-phase Mϕ. Inhibition of cyclic nucleotide phosphodiesterase type 4 (PDE4), which hydrolyzes cAMP, modulates disease progression in MRL/Mp–lpr/lpr lupus mice (70) and the PDE4 inhibitor apremilast is beneficial in human discoid lupus (71). It will be of interest in the future to see if PDE4 inhibitors promote expansion of the CD138⁺ Mϕ subset via CREB activation.

The authors have no financial conflicts of interest.