Toll-like receptor 7-stimulated tumor necrosis factor α causes bone marrow damage in systemic lupus erythematosus

Abstract

Objective

To define the pathogenesis of bone marrow (BM) involvement in systemic lupus erythematosus (SLE).

Methods

Tumor necrosis factor-α (TNFα), cell death, and cellular damage in BM from SLE patients, controls, and mice with pristane-induced lupus were analyzed morphologically and using immunohistochemistry. The pathogenesis of BM abnormalities was studied in wild-type, and TNFα-, TLR7-, and interferon-α receptor-deficient, along with B cell-deficient (µmt) mice treated with pristane. Flow cytometry was used to examine TNFα production (intracellular staining) and plasma cell/plasmablast development. CXCL12 expression was determined by quantitative PCR.

Results

SLE patients’ BM exhibited striking death of niche and hematopoietic cells associated with TNFα over-production. BM from mice with an IFN-I-mediated lupus syndrome induced by pristane showed similar abnormalities. TNFα was produced mainly by BM neutrophils, many with phagocytosed nuclear material (LE cells). TNFα production was abolished in TLR7−/− and µmt mice but was restored in µmt mice by infusing normal plasma. Pristane-treated wild-type- and IFNAR−/− mice developed anemia, BM hypocellularity, and extramedullary hematopoiesis, which were absent in TLR7−/− and TNFα−/− mice. Additionally, CXCL12, which is produced by stromal cells and mediates homing of hematopoietic cells and plasmablasts, was decreased in BM from pristane-treated wild-type mice but normal in TNFα−/− mice.

Conclusion

Although autoantibodies and glomerulonephritis are IFN-I dependent, lupus-associated BM abnormalities were TLR7- and TNFα-driven, but IFN-I-independent, suggesting that lupus is a disorder of innate immunity in which TLR7 activation by phagocytosed nuclei causes relentless IFN-I and TNFα production mediating glomerulonephritis and hematologic involvement, respectively.

Systemic lupus erythematosus (SLE) is a chronic multiorgan inflammatory disease in which autoantibodies against nucleic acid-protein complexes, such as chromatin and ribonucleoproteins, cause disease by forming immune complexes that deposit in target tissues (1). Hematological manifestations of SLE may be autoantibody-mediated or a consequence of renal insufficiency or inflammation. The pathogenesis of anemia of chronic inflammation, the most frequent hematological manifestation (2), is incompletely understood. Increased Type I interferon (IFN-I) levels, associated with renal, central nervous system, and hematological manifestations (3), may play a role. However, in rheumatoid arthritis (RA) (4, 5), anemia of chronic inflammation is thought to be tumor necrosis factor-α (TNFα)-mediated (6).

An IFN-I dependent lupus syndrome closely resembling human SLE develops in BALB/c, C57/BL6 (B6), and other strains of mice with chronic inflammation following i.p. pristane (2,6,10,14-tetramethylpentadecane, TMPD) injection (7). Autoantibody production and glomerulonephritis in TMPD-lupus require toll like receptor 7 (TLR7)-mediated IFN-I production driven by transcription factors IRF5 and IRF7 (8, 9). Here, we examine the bone marrow (BM) abnormalities in SLE patients and mice with TMPD-induced lupus to better define the pathogenesis of hematological dysfunction. TMPD-treated mice developed anemia and cell death in the BM, which were IFN-I-independent but TNFα-dependent. TLR7-stimulated TNFα production in the BM caused niche dysfunction, dyserythropoiesis, and anemia in TMPD-lupus. Similar abnormalities develop in SLE patients, suggesting that TNFα-mediated BM dysfunction also contributes to the hematological manifestations of human SLE.

Patients and Methods

Patients

Pathology records over the past 10 years from the University of Florida were reviewed and 11 BM aspirates/core biopsies with a diagnosis of SLE were identified. Patients with no core biopsy, insufficient tissue for accurate diagnosis, or insufficient clinical data to confirm a diagnosis of SLE were excluded. Six suitable patients were identified. Wright-Giemsa stained BM aspirate smears and cytospin preparations, hematoxylin and eosin (H&E)-stained and reticulin stained BM core biopsies were reviewed by a hematopathologist. Immunohistochemistry (IHC) for TNFα and cleaved caspase-3 was performed on core biopsies and expression levels were quantified by morphometric analysis (see below). SLE was classified using the ACR criteria (10). Normal BM specimens from individuals undergoing staging for neoplasms (mainly lymphomas) were selected as controls. Human studies were reviewed and approved by the UF IRB. The studies were performed using leftover/deidentified human tissue and were deemed not to require informed consent. Clinical data were analyzed by LY and WR. All authors had access to these data.

Mice

Mice were maintained under specific pathogen free conditions at the University of Florida Animal Facility. B6 TNFα−/− and B-cell-deficient (µmt) mice were from Jackson Laboratory (Bar Harbor, ME). BALB/c TLR7−/− mice, from Dr. Shizuo Akira, were obtained from Oriental Bioservices (Kyoto, Japan). Type I interferon receptor deficient (IFNAR−/−) mice backcrossed nine generations onto a BALB/c background, were provided by Dr. Joan Durbin (Nationwide Children’s Hospital, Ohio State University, Columbus OH). Wild-type BALB/cJ and B6 mice were from Jackson Laboratory. Mice received 0.5 mL of TMPD (Sigma, St. Louis, MO) i.p. or left untreated. BM cells were isolated from tibias and femurs by flushing with 3 ml PBS. Animal studies received prior approval by the UF IACUC and were conducted in compliance with its recommendations.

Histology and IHC

Mouse tibias/femurs and BM core biopsies from SLE patients and controls were fixed in 10% neutral buffered formalin for 1hr and decalcified in Rapid-Cal-Immuno-Decal Solution (BBC Biochemical, Stanwood, WA) for 2hr. Paraffin sections (4-µm) were stained with H&E. For IHC, paraffin sections were dried on slides for 2hr at 60°C. Slides were placed in a Ventana Medical Systems (Tucson, AZ) automated immunostainer and deparaffinized. Heat-induced epitope retrieval was performed with Ventana’s CC1 retrieval solution (30 min at 95–100°C). Primary antibodies anti-cleaved-caspase-3 (Cell Signaling, Danvers, MA), anti-TNFα (Abcam, Cambridge, MA), and anti-CD71 (Dako/Agilent Technologies, Santa Clara, CA) were applied for 32min at 37°C followed by peroxidase- or alkaline phosphatase-conjugated goat anti-mouse or goat anti-rabbit secondary antibodies (30min). Reaction product was visualized using Ultra View DAB (brown) or Alkaline Phosphatase Red detection kit (Ventana). Slides were counterstained with Ventana hematoxylin.

Computer-assisted histomorphometric analysis

TNFα and caspase-3 immunostained slides (n=3) for each case were viewed with an Olympus microscope interfaced with a digital camera. Representative images (6–12) from each slide were photographed (100X objective) and acquired with the Olympus CellSens standard photo image capture software (Olympus Global, Japan). The expression area and intensity of staining were quantified using MetaMorph Premier Image Analysis Software (Molecular Devices Corporation, Sunnyvale, CA). Staining intensity (thresholded area) was expressed as percentage of total examined BM cellular area after subtracting BM noncellular space (bone trabeculae and fat lobules) from total area.

Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assays

Four µM sections of paraffin-embedded human BM biopsies were deparaffinized and subjected to antigen retrieval. Cell death was visualized in situ using anti-digoxigenin-conjugate peroxidase-DAB (brown)-based colorimetric detection (ApopTag Peroxidase In Situ Apoptosis Detection Kit, Chemicon/Millipore, Danvers, MA). TUNEL-stained slides subsequently were incubated with anti-CD71 antibodies and reaction product was visualized by the Ultra View Alkaline Phosphatase Red detection kit. Mouse IHC was performed by the University of Florida Molecular Pathology and Immunology Core and IHC of human tissue was performed by the Shands Hospital Histology/IHC Laboratory.

Real-time quantitative PCR (Q-PCR

Q-PCR was performed as described (8). Briefly, total RNA was extracted from BM cells using TRIzol (Invitrogen, Carlsbad, CA) and cDNA was synthesized using the iScript™ cDNA Synthesis Kit (Bio-Rad, Hercules, CA). SYBR green QPCR was performed using an Opticon II thermocycler (Bio-Rad). Amplification conditions were: 95°C for 10min, followed by 45 cycles of 94°C for 15sec, 60°C for 25sec, and 72°C for 25sec. After the final extension (72°C for 10min), a melting-curve analysis was performed to ensure specificity of the products. Primer sequences were as follows: CXCL12 forward TGCTCTCTGCTTCCTCCA; CXCL12 reverse GGTCCGTCAGGCTACAGAGGT; TNFα forward AAGGGATGAGAAGTTCCCAAA; TNFα reverse CACTTGGTGGTTTGCTACGA.

Flow cytometry

BM cells, splenocytes, and peritoneal cells were stained with optimally diluted primary antibodies or isotype controls (30min 22°C), washed, and resuspended in PBS/0.5% BSA. Fifty thousand to 100,000 events per sample were acquired using a BD LSRII flow cytometer (BD Bioscience, San Jose, CA) and analyzed with FCS Express 3 (De Novo Software, Ontario, Canada). Anti-Ly6G, anti-B220, anti-CD11b, anti-B220, and anti-TNFα conjugated to allophycocyanin (APC), Pacific Blue (PB), fluorescein isothiocyanate (FITC), phycoerythrin (PE), or peridinin-chlorophyll-protein (PerCP) were obtained from eBioscience (San Diego, CA). Anti-CD138-APC and anti-CD19-PB were from BioLegend (San Diego, CA).

Intracellular cytokine staining

Intracellular TNFα was detected after culturing BM, peritoneal, or spleen cells (2.5 × 10⁶ cells/ml) in RPMI-1640/10% FBS and GolgiStop (5 µg/ml) at 37°C. Six hours later, cells were fixed (Fixation/Permeabilization buffer, eBioscience), surface stained, washed (Perm/Wash buffer, eBioscience), and intracellularly stained with anti-TNFα antibodies (BD Bioscience) or isotype controls before flow cytometry. Data were analyzed with FCS Express 3 software.

Plasma transfer

B6 mice received either 0.5 ml TMPD i.p. or left untreated. Two weeks later, 100µl of plasma from either TMPD-treated or untreated B6 mice was given via the tail vein to B6 µmT mice that either were treated 2-wks earlier with TMPD (0.5 ml i.p.) or left untreated. BM cells were isolated from tibias and femurs at day 5. Intracellular TNFα was detected as above.

Hematological testing

Five µl of mouse blood was diluted with 15µl RPMI-1640/5% heparin. Hemoglobin and counts (erythrocytes, platelets, leukocytes) were measured by Coulter Counter (BD Bioscience).

Statistics

Data are presented as mean ± SEM. Unless otherwise stated, comparisons between mean values were performed by ANOVA or 2-tailed Student’s t-test as appropriate using GraphPad (San Diego, CA) Prism version-5 software. A P value of <0.05 was considered significant.

Results

BM abnormalities in SLE patients

There are relatively few studies of the pathogenesis of anemia of chronic inflammation in lupus. We retrospectively examined histological changes, cell death, and TNFα production in SLE patients’ and control BM biopsies/aspirates (Table S1). Five of 6 SLE patients had nephritis and 3 were direct Coombs⁺ (one with hemolytic anemia). Medications included corticosteroids (5/6 patients) and mofetil mycophenolate or azathioprine (4/6 patients). Five of six BM biopsies exhibited hypocellularity and all showed erythroid dyspoiesis and reticulin fibrosis, which were uniformly absent in control BM biopsies (Table S2). Hemoglobin ranged from 7.4–10.3 (mean 9.1g/dL) in SLE patients and from 8.6–15.2 (mean 11.9g/dL) in controls. Mean WBC was 3800/mm³ in patients and 8200/mm³ in controls, and mean platelet counts were 133,000/mm³ and 258,000/mm³ respectively.

BM aspirates revealed numerous dead cells (Fig. S1A), erythroid dyspoiesis (Fig. S1B), plasmacytosis (Fig. S1C), hemophagocytosis (Fig. S1D), and phagocytosis of nuclear material by mature neutrophils (LE cells) (Fig. S1D, inset) or other cell types (Fig. S1B, inset). Numerous LE cells were seen in 4/6 SLE BM aspirates (Table S1). Additional abnormalities seen on BM core biopsies included hypocellularity, BM stromal damage/disorganization, interstitial lymphoid aggregates, and reticulin fibrosis (Fig. S1E–H) (Table S2).

Since histology revealed erythroid dyspoiesis with morphological evidence of cell death in all of the SLE patients (Table S1–2, Fig. S1), BM was stained with anti-cleaved caspase-3 antibodies, a specific marker of cell death (Fig. 1A). Numerous caspase-3⁺ cells were seen adjacent to bone trabeculae and in interstitial regions of SLE BM, whereas caspase-3⁺ cells were rare in control BM. Cell death also could be demonstrated in lupus BM by TUNEL, and co-staining with anti-CD71 antibodies revealed prominent death of erythroid precursors (Fig. 1B). Less extensive cell death of the myeloid and megakaryocytic lineages also was demonstrated (not shown). In SLE, caspase-3⁺ cells occupied 9–12% of the total BM area vs. 1–2% in control BM (Fig. 1C).

Figure 1. IHC of lupus BM.

A, IHC for cleaved caspase-3 (1:100, red staining) in SLE and control BM. Numerous elongated paratrabecular caspase-3⁺ cells were apparent (top) with smaller numbers in interstitial locations. Dotted lines indicate locations of bone trabeculae (BT). B, Double IHC showing death of CD71⁺ cells (red, CD71 staining, brown TUNEL) in SLE and control BM. Dual (brown/red) staining in the SLE BM biopsy indicates extensive death of erythroid cells. C, MetaMorph quantification of caspase-3 staining in BM from six SLE patients and six controls expressed as % of total area. Right, comparison of the SLE and Control groups (P < 0.0001, Student t-test). D, IHC for TNFα in interstitial portions of the BM. Note intense signal surrounding viable BM neutrophils (arrows). E, IHC (brown stain) for TNFα in SLE vs. control BM. Note intense staining near the bone trabeculae (BT) in SLE but not control BM. F, morphometric analysis (MetaMorph software) of TNFα IHC showing intense staining adjacent to the bone trabeculae (BT) and less extensive staining of interstitial regions (Int). G, MetaMorph quantification of TNFα staining in BM from SLE patients and controls expressed as % of total area. Right, comparison of the SLE and Control groups (P < 0.0001, Student t-test). H, Anti-caspase-3 IHC of BM from TMPD-treated B6 and B6-TNFα−/− mice. Red arrows, apoptotic/necrotic cells (brown stain); white arrows, viable cells. Right, quantification of caspase-3⁺ cells (Metamorph software).

As TNFα promotes Fas-mediated apoptosis and may damage hematopoietic precursors and/or stromal cells (4, 11, 12), we examined its production in SLE BM. Anti-TNFα antibodies exhibited both intracellular and extracellular staining of the BM. In some fields, intense staining could be seen surrounding neutrophils (Fig. 1D) and monocytes (not shown). The extent of TNFα staining was dramatically higher in SLE patients’ BM biopsies than in controls (Table S2). Staining was particularly intense adjacent to presumptive osteoblasts lining the surface of bone trabeculae (Fig. 1E,F), though there also was staining in interstitial regions (Fig. 1F). There was little TNFα staining of control BM (Fig. 1E). Ten-18% of the area in SLE BM was stained with anti-TNFα vs. 1–3.5% in controls (P < 0.0001, Student t-test) (Fig. 1F,G).

Taken together, these studies indicate that SLE patients undergoing BM biopsy share a number of histological/immunological features, including BM hypocellularity and erythroid dyspoiesis, presence of numerous LE cells, marked TNFα production, mild reticulin fibrosis, and death of hematopoietic niche cells and hematopoietic cells, especially erythroid precursors. To investigate the mechanisms, we looked for similar abnormalities in mice with experimental lupus.

Hematological abnormalities in TMPD-lupus

The TMPD-lupus model closely resembles human interferon-signature-associated SLE (7). H&E staining revealed that BM from TMPD-treated wild-type and IFNAR−/− mice was markedly hypocellular (Fig. 2A). In contrast, hypocellularity was absent in pristane-treated TLR7−/− or TNFα−/− mice. Consistent with our previous observations (13), TMPD-treated wild-type mice exhibited hepatosplenomegaly. Histological evaluation revealed extramedullary hematopoiesis (EMH) in the spleen and liver (Fig. 2B–C). EMH may develop as a compensatory mechanism in patients with chronic anemia (14). Although a low level of EMH is common in spleens of normal mice (14), its extent far exceeded what is normally seen. The organization of lymphoid nodules (white pulp) (Fig. 2B) in the spleen of TMPD-treated mice was disorganized and megakaryocytes were increased (arrows). Similarly, liver from TMPD-treated mice showed EMH involving the myeloid, erythroid, and megakaryocytic lineages (Fig. 2C). In parallel with decreased BM damage in TNFα−/− mice (Fig. 2A), EMH was less extensive in livers of TMPD-treated TNFα−/− mice (Fig. 2C). Consistent with the BM hypocellularity, total red blood cell count, hemoglobin, and hematocrit all were reduced in TMPD-treated mice vs. untreated controls (Table 1). These abnormalities were absent in TNFα−/− mice. In contrast, TMPD did not alter leukocyte or platelet counts in wild-type mice (Table 1).

Figure 2. Histology of the BM and spleen of TMPD treated mice.

A, H&E staining of femurs or tibias from BALB/c (WT), IFNAR−/−, TLR7−/−, and TNFα−/− mice treated with TMPD or left untreated (No Tx). B, H&E staining showing EMH in spleen of BALB/c mice treated with TMPD. Arrows, megakaryocytes; Fo, B cell follicle. C, H&E staining of EMH in liver from B6 or B6 TNFα−/− mice treated with TMPD or left untreated showing myeloid, erythroid, or megakaryocytic (Meg, arrow) precursors.

Table 1.

Anemia Induced by TMPD Is TNFα-dependent

	B6			B6 TNFα−/−
	No Tx (n = 8)	TMPD (n = 8)	P-value*	No Tx (n = 7)	TMPD (n = 7)	P-value*
RBC (X 106/µl)	13.1 ± 2.53	7.95 ± 0.87	< 0.001	10.8 ± 3.35	8.60 ± 0.87 NS
Hemoglobin (g/dL)	17.6 ± 3.06	12.4 ± 1.32	< 0.01	16.0 ± 2.47	13.6 ± 1.64	NS
Hematocrit (%)	55.5 ± 4.39	38.7 ± 3.82	< 0.01	50.5 ± 5.16	43.6 ± 7.7	NS
Platelets (X 106/ml)	374 ± 96.8	403 ± 71.7	NS	388 ± 121	371 ± 61.5	NS
WBC (X 106/ml)	12.1 ± 2.06	10.4 ± 0.87	NS	11.6 ± 4.34	14.9 ± 3.57	NS

^*Two-way ANOVA; No Tx, untreated controls

As in SLE patients, extensive cell death was apparent in BM from TMPD-treated wild-type mice by caspase-3 staining (Fig. 1H) and annexin V/7AAD staining (flow cytometry) (not shown). These similarities suggested that, as in human SLE, TNFα also might be produced in the BM of mice with TMPD-lupus, prompting us to perform intracellular staining for TNFα.

TNFα is produced by BM neutrophils and monocytes and is TLR7-dependent

We investigated TNFα production in BM from B6 mice 2wks after TMPD injection (Fig. 3). BM cells were cultured with GolgiStop before surface staining and intracellular TNFα staining. Viable intracellular TNFα⁺ BM cells were visualized readily after 4–6hr culture (Fig. 3A). TNFα was produced exclusively by CD11b⁺CD3⁻B220⁻ cells, mostly Ly6G⁺ neutrophils, but also TNFα⁺CD11b⁺Ly6G⁻ monocytes/macrophages. Ly6G⁺TNFα⁺ cells also were found in the peritoneum, but were absent in spleen (Fig. 3B), indicating that TMPD-stimulated TNFα production was a local phenomenon.

Figure 3. Intracellular TNFα production (flow cytometry.

A, BM cells from TMPDtreated B6 mice were incubated with GolgiStop for 4 or 6-hours before staining with anti-TNFα-APC, anti-CD11b-PE, anti-Ly6G-PB, anti-CD3-FITC, and anti-B220-PerCP antibodies. TNFα producing cells were CD11b⁺Ly6G⁺CD3⁻B220⁻. B, TNFα production by TMPD-treated B6 peritoneal exudate cells (PEC), BM cells, and splenocytes (spleen). Ly6G⁺TNFα⁺ cells were found in BM and peritoneum but not spleen. C, Intracellular TNFα+Ly6G staining in BM and PEC from TMPD-treated BALB/c, BALB/c IFNAR−/−, and BALB/c TLR7−/− mice or untreated controls (No tx). D, Percentage of Ly6G⁺TNFα⁺ cells in the BM of BALB/c vs. IFNAR−/− and TLR7−/− mice. (* P < 0.05; *** P < 0.001, 2-way ANOVA; n.s., not significant).

Induction of autoantibodies and nephritis by TMPD is abolished in TLR7−/− and IFNAR−/− mice (7). Unexpectedly, IFNAR-deficiency had little effect on BM hypocellularity (Fig. 2A). In wild-type mice the BM and peritoneum both contained intracellular TNFα⁺Ly6G⁺ neutrophils as well as TNFα⁺Ly6G⁻ monocyte/macrophages, although in BM TNFα was seen mainly in neutrophils (Fig. 3B). TNFα-producing cells also were found in TMPD-treated IFNAR−/− mice (Fig. 3C–D). In contrast, although CD11b⁺Ly6G⁺ neutrophils and CD11b⁺Ly6G⁻ monocytes were present in the BM and peritoneum of TLR7−/− mice, TNFα producing cells were absent (Fig. 3C–D). Thus, TMPD-stimulated TNFα production required TLR7 but not signaling through the IFNAR.

TNFα production by BM neutrophils requires B cells

TNFα expression in BM, but not spleen, following i.p. TMPD injection led us to examine whether circulating immunoglobulin was involved. BM from TMPD-treated wild-type, but not µmt, mice contained TNFα⁺ neutrophils (Fig. 4A). To evaluate whether this was due to immunoglobulin secretion, µmt mice were treated i.p. with TMPD followed 2wks later by i.v. injection of untreated wild-type mouse plasma or plasma from a wild-type mouse treated 2wks earlier with TMPD (Fig. 4B,C). TNFα⁺ cells were seen in TMPD-treated µmt mice following plasma transfer from either TMPD-treated or untreated B6 donors. In untreated µmt mice, no TNFα was produced following plasma transfer from untreated B6 donors. However, when plasma from TMPD-treated B6 donors was transferred to untreated µmt mice, the BM contained small numbers of TNFα⁺ neutrophils.

Figure 4. Role of B cells in TMPD-stimulated TNFα production.

A, Flow cytometry of Ly6G⁺TNFα⁺ BM cells in TMPD-treated B6 µmt mice and B6 controls (6/group). B, Flow cytometry of Ly6G⁺TNFα⁺ BM cells in TMPD-treated or untreated (No Tx) B6 µmt mice infused with plasma from untreated or TMPD-treated wild-type (WT) B6 mice. Additional control mice did not receive any plasma (No plasma). Representative of ≥6 experiments. C, Percentage of Ly6G⁺TNFα⁺ cells in the BM of µmt mice with/without plasma transferred from WT mice. D, BM CXCL12 expression (Q-PCR). E, Spleen CXCL12 expression. F, BM PCs and PBs. BM cells (gated on live cells) from TMPD-treated mice or untreated controls were stained with anti-CD19-PB plus anti-CD138-APC. Percentages of PCs (CD19^loCD138^hitop right) and PBs (CD19^+/−CD138^int, bottom right) are shown (* P < 0.05; ** P < 0.01; *** P < 0.001, 2-way ANOVA; n.s., not significant).

Local TNFα production suppresses BM CXCL12

Hematopoietic stem cells are regulated by specialized microenvironments (“niches”) in the BM, one of which is formed by osteoblasts lining the bone surface (15). Other niches are localized near endothelial cells. CXCL12, a chemokine produced by mesenchymal-derived cells plays an important role supporting hematopoiesis and the survival of long-lived plasma cells in the BM (16, 17). CXCR4/CXCL12 signaling promotes accumulation of plasma cells (PC) in the BM and the CXCR4 antagonist AMD3100 depletes myeloma cells from human BM (18). BM in TMPD-lupus contains few PCs/plasmablasts (PB) and low levels of the PC/PB attractive chemokine CXCL12 (19).

Intraperitoneal TMPD dramatically suppressed BM CXCL12 mRNA expression in wild-type mice (Fig. 4D). CXCL12 expression also was suppressed in IFNAR−/−, but not TNFα−/−, mice. In contrast, CXCL12 expression in the spleen was enhanced in TMPD-treated mice (Fig. 4E) and this was unaffected by TNFα or IFNAR deficiency. Consistent with the low CXCL12 expression in BM of TMPD-treated mice, the percentage of CD19⁺CD138^int PB in the BM of wild-type mice was decreased in comparison with untreated controls (Fig. 4F). Percentages of CD19⁺CD138^int PB in BM were unchanged by TMPD treatment in TNFα−/− and TLR7−/− mice, whereas they decreased in IFNAR−/− mice. In contrast, the percentage of more mature (CD19^+/− CD138^hi) PC in wild-type BM and also in BM from IFNAR, TLR7, or TNFα deficient mice was unchanged by TMPD treatment (Fig. 4F).

Discussion

Relatively little is known about the pathogenesis of non-autoantibody-mediated hematological manifestations of lupus. We defined several abnormalities characteristic of the BM of SLE patients and mice with TMPD-lupus, including BM hypocellularity/dyspoiesis, LE cells, extensive cell death, and BM stromal cell “niche” dysfunction (reticulin fibrosis and death of niche cells lining bone trabeculae in humans and decreased CXCL12 and PB in mice). Unexpectedly, in pristane-lupus these changes were mediated by TLR7-dependent local (BM) TNFα production rather than dysregulated IFN-I expression.

Dyserythropoiesis and anemia due to TLR7-driven TNFα production

Anemia of chronic disease or leukopenia is present in ~75% of SLE patients (2) and BM apoptosis has been noted morphologically (20). Reversible dyserythropoiesis is associated with active SLE (21) and BM stromal cell dysfunction (22), cell death (90%), hypocellularity (57.5%), and dyserythropoiesis (100%) all have been reported previously in SLE BM (23). The present study extends these observations by presenting quantitative evidence of extensive cell death and TNFα production in 100% of lupus BM biopsies. The retrospective design is a limitation this study, since only patients with the most severe hematological changes generally undergo BM analysis and further study will be necessary to determine whether patients with milder disease exhibit similar changes.

Our studies provide the first evidence of inflammation-induced hematological involvement in murine lupus and suggest that pristane-lupus is a suitable model for hematological lupus. The absence of hypocellularity, dyserythropoiesis, and anemia in TNFα−/− and TLR7−/− mice (Fig. 2, Table 1) indicates that in contrast to nephritis and autoantibody production (7), hematological involvement is due to TLR7-driven TNFα production. Interestingly, the mice developed anemia without leukopenia or thrombocytopenia and in SLE patients’ BM cell death was most pronounced in the erythroid lineage (Fig. 1B). TNFα production may selectively damage erythroid precursors, leading to anemia as suggested in RA (4). Thus, TMPD-lupus is mediated by multiple cytokines: although IFN-I is central to nephritis and autoantibody production, hematological involvement is TNFα-dependent (Fig. 5). Arthritis in TMPD-treated mice responds to TNFα-inhibitor therapy (24), suggesting that it also is TNFα-mediated. Interestingly, arthritis (and renal disease) in SLE patients may respond to TNFα inhibitor therapy (25) and recent studies suggest that TNFα is elevated in many SLE patients (26).

Figure 5. Model of innate immune dysfunction in lupus BM.

Apoptotic/necrotic cells are cleared by monocytes/macrophages (MΦ) expressing receptors that inhibit the inflammatory response, such as TAM receptors. Abnormal phagocyte function may allow opsonization by immunoglobulin (Ig) and uptake by Fc receptors, with transport of endogenous RNA ligands to endosomes containing TLR7. In monocyte/MΦ, TLR7 engagement promotes IFN-I production preferentially over TNFα, whereas neutrophils preferentially make TNFα. In TMPD-lupus, autoantibody production and lupus nephritis are IFN-I-mediated, whereas anemia, niche damage, and arthritis are TNFα-mediated. The spectrum of clinical manifestations in lupus may reflect the balance of IFN-I and TNFα produced downstream of TLR7.

At present, we cannot determine the relative importance of TNFα vs. IFN-I in the hematological manifestations of human SLE. The interferon signature is associated with severe leukopenia, thrombocytopenia, and hemolytic anemia, manifestations not seen in TMPD-lupus (3). But its relationship to anemia of chronic disease has not been examined systematically. Our studies suggest that local TNFα production in SLE patients could damage hematopoietic precursors or stromal cells, as in myelodysplastic syndromes and RA (4, 12). Also, TNFα upregulates Fas expression in BM CD34⁺ cells, promoting Fas-mediated apoptosis (11) and a neutralizing anti-TNFα mAb improves anemia and decreases apoptotic erythroblasts in human TNFα transgenic mice (27). Consistent with this model, Fas/FasL-deficient mice are resistant to TMPD-lupus (28). Alternatively dyserythropoiesis may reflect decreased CXCL12 production by BM stromal cells, consistent with previous observations that TNFα inhibits CXCL12 (29, 30) and the importance of CXCL12 in hematopoiesis.

BM stromal cell dysfunction in lupus

Osteoblasts and their mesenchymal precursors (mesenchymal stem cells) line the bone surface forming niches that promote hematopoiesis. Administration of parathyroid hormone simultaneously increases both osteoblasts and hematopoietic stem cells (HSCs) in mice (15). Cell death in paratrabecular areas of SLE BM (Fig. 1A) and its close relationship to TNFα production (Fig. 1E,F) suggest that the loss of these and perhaps other niche cells may contribute to the cytopenias in SLE. The TNFα-dependent reduction of BM CXCL12 production in TMPD-lupus (Fig. 4) provides further evidence that BM stromal cell dysfunction may be important clinically.

CXCL12-CXCR4 interactions regulate the migration and recruitment of HSCs and PCs to the BM (18, 31). CXCL12 is produced in BM by mesenchymal stem cells, osteoblasts, endothelial cells (32) and other cell types, and helps establish PC survival niches (16, 17). In mice, decreased BM CXCL12 was associated with low numbers of BM PBs, but not PCs (Fig. 4E). Both were restored to control levels in TNFα−/− or TLR7−/− mice, suggesting that TMPD treatment causes BM stromal cell dysfunction in mice, consistent with a report that lupus stromal cells poorly support hematopoiesis (33).

TLR7-driven TNFα production in BM neutrophils

TMPD caused TLR7-dependent TNFα production by CD11b⁺Ly6G⁺Ly6C^int BM neutrophils and to a lesser degree by monocytes. Neutrophils express TLR7 and respond to TLR7 ligands by producing cytokines (34, 35). TLR7 and TLR9 are expressed only by the PMN-II (CD49⁻CD11b⁺) neutrophil subset, which produces IL-10, TNFα, and IL-1β and contains a ring-shaped nucleus in mice (35). Neutrophils with this morphology predominate in peritoneal exudates of TMPD-treated mice (36). The PMN-I subset (CD49⁺CD11b⁻) produces IL-12 and TNFα, but does not express TLR7 (35). Thus, the CD11b⁺TLR7-responsive BM neutrophils producing TNFα in TMPD-lupus may be PMN-II.

The rate of cell death in TMPD-treated mice may exceed the ability of BM phagocytes (monocytes/macrophages) to clear the debris (37) and/or the presence of TMPD in the BM might inhibit phagocytosis (38). Dead cells in lupus BM represent an abundant source of endogenous TLR7 ligands (39, 40) that may be taken up by neutrophils (LE cells, Fig. S1D), leading to TLR7 activation (Fig. 5). Normally, the uptake of apoptotic cells by phagocytes is non-inflammatory (41, 42) and mediated by TAM receptors or other non-inflammatory receptors (43). But abnormal phagocyte function in lupus (44) may allow opsonization of apoptotic/necrotic cells, uptake via Fc receptors, and engagement of endosomal TLR7 by endogenous TLR7 ligands (45). Consistent with this model, TMPD-stimulated BM neutrophil TNFα production was TLR7-dependent, abolished in B cell-deficient mice, and restored by infusing plasma (Fig. 4). Immunoglobulin and/or complement may opsonize nuclear material released in the BM leading to uptake via Fcγ receptors (FcγRI, FcγRIII, FcγRIV), Fcµ receptors, or complement receptors (41, 46). However, it remains to be determined precisely how peritoneal inflammation causes BM neutrophils to produce TNFα.

EMH in TMPD-treated mice

Due to EMH, TMPD treated mice exhibit marked hepatosplenomegaly (13) (H Zhuang, unpublished data). EMH (Fig. 2) may partially compensate for BM dyspoiesis in TMPD-lupus and the numerous megakaryocytes (Fig. 2) may help maintain normal platelet counts in TMPD-lupus. In humans, EMH is a feature of myeloprolferative disorders associated with the JAK2^V617F mutation, which causes overproduction of TNFα, suppressing hematopoiesis (47, 48). It will be of interest to investigate whether the JAK2 pathway (47) is involved in TNFα production and niche cell death in SLE. All SLE patients studied here showed mild-moderate reticulin myelofibrosis (Table S2) and myelofibrosis/EMH has been reported in human SLE (49). Finally, splenomegaly and EMH are prominent features of lupus-like disease in mice lacking the ubiquitin-editing protein A20 (Tnfaip3) in dendritic cells (50). As A20 suppresses NFκB-mediated proinflammatory cytokine production, the lupus phenotype of A20-deficient mice supports the idea that certain clinical manifestations of SLE may reflect TNFα production downstream of TLR7.

In summary, TMPD-lupus, a model of human SLE associated with the interferon signature, is a disease of innate immunity involving abnormal regulation of TLR7 signaling and production of TNFα as well as IFN-I (Fig. 5). The importance of dysregulated IFN-I production in the pathogenesis of SLE is clear in both humans (3) and mice (7). The present data suggest that TNFα production downstream of TLR7 also plays a role in the disease, consistent with the high serum levels of TNFα (26) and efficacy of TNFα inhibitor therapy in some SLE patients (25).