Research ArticleIMMUNOLOGY

Critical role of synovial tissue–resident macrophage niche in joint homeostasis and suppression of chronic inflammation

See allHide authors and affiliations

Science Advances  06 Jan 2021:
Vol. 7, no. 2, eabd0515
DOI: 10.1126/sciadv.abd0515


Little is known about the mechanisms regulating the transition of circulating monocytes into pro- or anti-inflammatory macrophages in chronic inflammation. Here, we took advantage of our novel mouse model of rheumatoid arthritis, in which Flip is deleted under the control of a CD11c promoter (HUPO mice). During synovial tissue homeostasis, both monocyte-derived F4/80int and self-renewing F4/80hi tissue–resident, macrophage populations were identified. However, in HUPO mice, decreased synovial tissue–resident macrophages preceded chronic arthritis, opened a niche permitting the influx of activated monocytes, with impaired ability to differentiate into F4/80hi tissue–resident macrophages. In contrast, Flip-replete monocytes entered the vacated niche and differentiated into tissue-resident macrophages, which suppressed arthritis. Genes important in macrophage tissue residency were reduced in HUPO F4/80hi macrophages and in leukocyte-rich rheumatoid arthritis synovial tissue monocytes. Our observations demonstrate that the macrophage tissue–resident niche is necessary for suppression of chronic inflammation and may contribute to the pathogenesis of rheumatoid arthritis.


Rheumatoid arthritis (RA) is a chronic inflammatory, erosive disease. Patients frequently develop circulating antibodies to citrullinated peptides (anti-CCP) before disease onset, known as preclinical RA. In the synovial tissue (ST) of patients with established RA, but not that of normal controls, anti-CCP antibodies are produced and citrullinated proteins are present (1). A major gap in our understanding of the pathogenesis of RA is identifying the mechanism by which systemic autoimmunity targets the ST. In RA ST chronic inflammation is characterized by increased neovascularization and infiltration with fibroblasts, B and T lymphocytes, dendritic cells, and macrophages (2). However, the number of sublining macrophages, but not other cell types, in RA ST is an important biomarker of RA disease activity, prognosis, and response to therapy (3).

cFLARE [also called FLIP (FLICE-like inhibitory protein)] has been identified as a risk locus in RA (4), and FLIP is highly expressed in RA ST macrophages (STMs) (5). Since FLIP is essential for preventing macrophage FasL-mediated apoptosis and thus a potential therapeutic target, we deleted Flip in myeloid cells, driven by a LysM Cre promoter (6). In contrast to expectations, these mice developed neutrophilia and monocytosis, with a marked reduction of macrophages in the spleen and lymph nodes, together with massive neutrophil infiltration, mediated by granulocyte colony-stimulating factor (7), and they died prematurely before 5 months of age (6). We further demonstrated that the deletion of Flip mediated by Cre recombinase under the control of a CD11c promoter (aka HUPO mice) results in less intense neutrophilia and monocytosis, with little or no reduction of macrophages in the spleen and lymph nodes (8). The HUPO mice did not die prematurely but unexpectedly developed a chronic, erosive arthritis, with the incidence increasing with age to 80% by ~22 weeks (8). Macrophages are critical for the initiation of this arthritis, since HUPO mice crossed to Rag−/− mice still develop a mild arthritis, because of infiltration with macrophages, while autoreactive T cells and autoantibodies are critical for disease progression (8). Therefore, we characterized macrophages in age-matched, littermate control and HUPO mice with or without arthritis. Positing that pathogenic macrophages in HUPO joints derive from circulating monocytes, we used a previously described gaiting strategy used to track monocytes entering into tissues (9, 10).

We identified F4/80hiMHCII (FH1) macrophages, in which Flip was highly expressed, as the dominant tissue-resident macrophage subset, which was essential during homeostasis for maintaining niche competency, preventing monocyte to tissue-resident macrophage differentiation. In contrast, in HUPO mice, Flip was reduced in F4/80hi macrophages, associated with reduction of the FH1 subset, opening a niche, permitting the influx of circulating proinflammatory monocytes, which differentiated into the F4/80hiMHCII+ (FH2) subset. In HUPO mice, increased apoptosis did not account for the reduction of the FH1 subset; however, reduced expression of receptors for efferocytosis, such as CD206 and CD163, and the reduction of apoptotic cells in the HUPO joints implicate reduced differentiation, which was directly observed following the phagocytosis of lysosome in vivo. Supporting its protective role, the FH1 subset was reduced before arthritis onset and inversely correlated with arthritis severity and duration. Bone marrow (BM) chimera and parabiosis experiments supported an important role of Flip and the chronic inflammatory environment in the differentiation of monocyte-derived FH2 into FH1 ST-resident macrophages, capable of suppressing chronic inflammation. Genes important in macrophage tissue residency are reduced in monocytes from inflammatory RA ST, suggesting a role for the disruption of the ST-resident macrophage niche in the transition from preclinical to clinical RA and the progression of disease.


STM subsets during homeostasis

To define STMs, we used an established gating strategy (9, 10), subsetting the CD11b+ population by expression of Ly6C, MHCII, and F4/80 (fig. S1A). ST F4/80+ macrophages were separated into those that were F4/80int (FI) and F4/80hi (FH). The FI population was then gated into three subsets based on Ly6C and MHCII expression (FI1, FI2, and FI3). We defined the Ly6C+ FI1 and FI2 subsets as STMs to distinguish them from true monocytes in the circulation, BM, and spleen (11). Supporting this distinction, clear differences in the expression of CD64, F4/80, and MHCII were noted between classical monocytes (CM) and FI1 and FI2 macrophages, although the mean fluorescence intensity of Ly6C was similar (fig. S1B). The F4/80hi population was further characterized as MHCII (FH1) and MHCII+ (FH2). Peripheral blood classical and non–classical monocytes (NCM) were identified by subsetting CD11b+CD115+ cells by Ly6C and CD62L (fig. S1C).

Distinct patterns of gene expression across subsets of STMs

We next examined transcriptional profiles by RNA sequencing (RNA-seq). The control FI2 subset was omitted because of low cell numbers. We defined 7778 genes as expressed in this dataset. Principal components analysis (PCA) suggested that each subset represented a distinct population of myeloid cells with CM and NCM from the blood clearly segregating (Fig. 1A). Pairwise correlation between subsets demonstrated a strong association between CM and NCM and a modest correlation of the FI1 synovial macrophage subset and CM (Fig. 1B), consistent with the notion that this population differentiated from CM (12). Notably, the FH1 subset demonstrates arguably the most distinct expression profile, in that it does not highly correlate to any other subset.

Fig. 1 Distinct patterns of gene expression across subsets of STMs during homeostasis determined by RNA-seq.

(A and B) PCA and pairwise Pearson’s correlation coefficient of gene expression (total 7778 genes expressed) from wild-type mice, across individual samples from blood CM and NCM and STM subsets (FI1, FI3, FH1, and FH2). (C) K-means clustering of 1936 differentially expressed genes (DEGs) across CM and the STM subsets. (D) Representative examples of RNA expression from genes identified in each of the six clusters in (C), presented as the means ± 1 SE in fragments per kilobase per million (FPKM).

To identify specific patterns of gene expression shared across STM subsets and CM, we performed unsupervised k-means clustering of 1936 genes that were differential across these populations (Fig. 1C and data file S1). We included CM in the clustering to highlight gene sets that were specific to the STMs and because CM migrate into tissues under homeostatic conditions (9, 10). We identified six distinct clusters of genes that were predominantly expressed in CM (cluster I), in FI1 (cluster II), or FI3 (cluster III) or shared to varying degrees across FI1, FI3, FH2, and FH1 macrophages (clusters IV to VI). The distinct set of genes associated with each cluster suggests differential functions, ontogeny, and degree of differentiation (Fig. 1, C and D). Genes in cluster VI are expressed in all STM subsets, but not CM, and include generic macrophage functions such as inflammatory response and metabolic processes. In contrast, cluster IV peaks in the FH1 subset and contains genes associated with tissue-specific functions, such as Cfs1r and Il10rb, which are involved with the maintenance of mature macrophages (13). This is reinforced by the expression of key macrophage maturity genes in cluster IV, such as Lamp1, Rab7, and Vamp3, important for lysosomal function and autophagy, and Msr1 and CD68, scavenger receptors important for maintaining homeostasis.

STMs are increased in HUPO mice, while the FH1 subset is reduced

Next, we examined STMs in HUPO mice and age-matched, littermate controls, at 22 to 37 weeks of age, when arthritis progression stabilized (8). The majority of F4/80+ macrophages of control mice were in the synovial lining, and they were increased in the both the lining and sublining of HUPO mice with arthritis (Fig. 2A). The patterns of the populations of STMs defined by flow cytometry were distinct between HUPO mice with or without arthritis and age-matched littermate controls (Fig. 2B). The total number of CD11b+CD64+F4/80+ macrophages was increased in HUPO mice with arthritis, and the percentages of the FI1 through FH2 subsets were variably increased in HUPO mice with arthritis (Fig. 2C). In contrast, FH1 STMs were significantly reduced in HUPO mice with arthritis, compared to those without arthritis or controls, and this subset inversely correlated with arthritis score and duration, while granulocytes, but not B cells, were positively associated (fig. S2A). The expression of MHCII was higher and F4/80 was lower on HUPO FH1, compared with control macrophages (fig. S2B), suggesting reduced differentiation of the HUPO FH1 macrophages that were present. Although evaluation of every tissue was not performed, CD11c+ macrophages were also reduced in the lungs of HUPO mice (fig. S2C).

Fig. 2 Altered histology and STM subsets during chronic arthritis and before disease onset in HUPO mice.

(A) Immunohistochemistry of ankle joints using anti-F4/80 or control immunoglobulin G for control or HUPO mice with arthritis. Right panel presents number of F4/80+ cells/0.01 mm2 and average lining thickness (n = 5 to 6 per group). (B) Representative flow cytometry of STMs from control and HUPO mice. The cells in the lower left quadrant of F4/80int macrophages were included when calculating the percent of each subset. (C) Number of total CD11b+CD64+F4/80+ STMs (left) and frequency of each subset in control and HUPO with or without arthritis mice (age, 22 to 37 weeks; n = 10 to 29). (D) The number of FH1, total, and FH2 STMs for mice (4 to 16 weeks of age) comparing control and HUPO mice with or without arthritis identified between 1 to 2 and 4 to 6 weeks (n = 5 to 17 per group). (E) Clinical scores and (F) histologic examination for young mice without or with arthritis. B, bone; C, cartilage; arrows identify synovial lining, and brackets identify the sublining in (A) and (F). Statistical analyses were performed by Student’s two-tailed t test for (A) and one-way analysis of variance (ANOVA) plus Tukey. *P < 0.05, **P < 0.01, and ***P < 0.001 among indicated groups. I indicates inflammation and C indicates cartilage destruction in (F).

The FH1 subset was reduced in young (4 to 16 weeks) HUPO mice before the onset of arthritis on clinical exam and before a significant increase of total STMs or the FH2 subset (Fig. 2D). Further, on histologic exam, low levels of inflammation and joint damage were observed before the onset of clinical arthritis, similar to preclinical or early RA (14), which increased over time, as arthritis severity increased (Fig. 2, E and F). These observations suggest that the reduction of the FH1 subset in HUPO mice may be associated with the development of arthritis. Of potential relevance to the increase of HUPO arthritis incidence with age (peak week, ~22), under homeostatic conditions, STMs decreased with age, and the FH1:FH2 ratio was (P < 0.01) reduced in older mice (fig. S2D).

Targeting Flip in STMs did not alter the apoptosis and proliferation of FH subsets during inflammation

Studies were performed to identify the factors contributing to the decrease of the FH1 and increase of the FH2, subsets in HUPO mice. Flip was low in monocytes from both HUPO and control mice (fig. S3A), consistent with human monocytes (15). Flip expression in ankle F4/80int (FI1 to FI3) macrophages was similar to monocytes in both HUPO and controls but was increased in both F4/80hi (FH1 and FH2) subsets of the control but not HUPO mice. Consistent with this observation, endogenous CD11c expression was low in HUPO CM through FI3 and increased in the F4/80hi subsets (fig. S3B). The variability of CD11c observed in the control FI1 and FI2 subsets may be related to increased apoptosis compared with those from HUPO mice (fig. S3, B and C). Although Flip was reduced, apoptosis of HUPO FH2-FH1 macrophages was not increased compared with controls (fig. S3C). The ability of the HUPO F4/80hi populations to survive may be related to an increase of anti-apoptotic proteins, such as Mif and Spp1 (fig. S3D), which protect against macrophage apoptosis (16, 17). Although our data did not address the mechanism for the reduction of FH1 macrophages in HUPO mice before arthritis onset, these observations do not support increased apoptosis of either F4/80hi subset for the ongoing reduction of the FH1 HUPO subset. Further, using 5-bromo-2′-deoxyuridine (BrdU) incorporation assessed 30 min after intravenous injection to determine whether the HUPO FH2 subset was expanding by local proliferation, we noted instead that the control FH2 subset was slowly proliferating during homeostasis compared with HUPO mice (fig. S3E) or with rapidly proliferating bone marrow monocyte (BMM) precursors (fig. S3F), as described (18). Therefore, increased proliferation was not responsible for the accumulation of the FH2 subset in this model of chronic arthritis.

Circulating monocytes are the major source of macrophages during chronic inflammation but not during homeostasis

Since increased proliferation did not account for the buildup of the HUPO FH2 subset, we examined the contribution of migration of monocytes by injecting BrdU on a single day, assessing the percentage of BrdU-positive monocytes in the BM, circulation, and STMs over time. On day 1, approximately 80% of the Ly6C+ BMM were BrdU-positive from control and HUPO mice (Fig. 3, A and B). At the same time, ~60% of CM, but few NCM, were labeled. The subsequent reduction of BrdU-positive CM on day 5 and the peak of BrdU-positive NCM on days 3 and 5 (~20 to 50%) are consistent with the short life span of CM and the differentiation of these cells into NCM, as previously described (12, 18). While the temporal patterns were similar, in separate analyses, HUPO BrdU-positive CM were higher than controls (P < 0.05) on days 1 and 3, and HUPO NCM were higher on day 3 (P < 0.001). In addition, injection of liposomes loaded with the fluorescent dye Dil [Dil-liposomes (Dil-lip)] resulted in comparable labeling of HUPO and control CM at 6 hours, which were both reduced at 24 hours (Fig. 3C). NCM incorporated Dil-lip to a greater degree, which was modestly reduced at 24 hours. Together, the observations with BrdU and Dil-lip do not suggest a major difference in the transition of CM to NCM between HUPO and control mice.

Fig. 3 Circulating monocytes are the major source of macrophages during chronic inflammation but not during homeostasis.

(A, B, D, and E) In vivo BrdU incorporation in control and HUPO mice, analyzed 1 to 5 days after BrdU administration. Values are the percent of BrdU-positive cells on the indicated days identified in BMMs (n = 3 to 5); circulating CM and NCM (n = 5 to 11); and five STM subsets (n = 6 to 11). (C) Uptake of fluorescent Dil-lip by CM and NCM of control and HUPO mice 6 and 24 hours after intravenous injection. (F) In vivo monocyte and STM depletion 24 hours following clodronate-loaded liposomes (Clo-lip) intravenous administration in control mice. Results are representative of two independent experiments. (G) Depletion of CM and NCM and (H) each subset of STMs in HUPO mice following treatment with Clo-lip (n = 7 to 8) or phosphate-buffered saline (PBS) (n = 4 to 5). Statistical analysis was performed by one-way ANOVA plus Tukey (A, B, D, and E) and by Student’s two-tailed t test (C, G, and H). Individual values and/or the means ± 1 SE are presented. *P < 0.05, **P < 0.01, and ***P < 0.001 among indicated groups. Fluorochromes: PE, phycoerthrin; APC, allophycoerthrin.

FI1 and FI2 macrophages exhibited high levels (~40 to 60%) of BrdU incorporation on days 1 and 3 in control and HUPO mice (Fig. 3, D and E), consistent their origin from circulating or BM-derived monocytes. Further, BrdU incorporation in the FI3 population peaked at 20 to 50% on day 3, suggesting a direct contribution from circulating MHCII+ NCM or differentiation from the FI1 and FI2 subsets. In contrast to the controls, which demonstrated no increase in the percentage of BrdU-positive cells in the FH1 and FH2 subsets between days 1 and 5, the BrdU-positive macrophages in these HUPO subsets were significantly increased on days 3 and 5, consistent with differentiation of BrdU-positive monocytes into FH macrophages. The percent HUPO FH1 BrdU-positive macrophages may have been relatively high on days 3 to 5 because of the low numbers of the FH1 subset; on those days, FH2 macrophages were numerically 5- to 14-fold greater than the FH1 subset. In control mice, clodronate-loaded liposomes effectively depleted monocytes and FI1 macrophages, but not the FH1 or FH2 subsets (Fig. 3F). In contrast, no significant reduction of HUPO CM or FI1 or FI2 macrophages occurred following treatment with clodronate-loaded liposomes (Fig. 3, G and H). Nevertheless, MHCII+ NCM and FI3 and FH2 macrophages were significantly (P < 0.05 to 0.01) reduced on day 2 or 3 following treatment with clodronate liposomes, consistent with the increased uptake of liposome by NCM (Fig. 3C). Although not definitively documented, these observations suggest that HUPO NCM may be entering the inflamed joints differentiating directly into FI3 and FH2 macrophages. Together, these data demonstrate that, in contrast to control mice, HUPO monocytes readily differentiate into F4/80hi STMs, although the FH1 subset remained reduced.

HUPO monocytes exhibit proinflammatory potential

We next examined circulating monocytes to better understand how they might contribute to HUPO arthritis. The total number of monocytes, specifically the CM, were significantly (P < 0.001) increased in HUPO mice with arthritis, and the expression of MHCII was increased (P < 0.001) on HUPO CM and NCM (fig. S4A). Of the 4480 genes expressed in either HUPO or control CM or NCM, 150 differentially expressed genes (DEGs) were up-regulated in both populations in HUPO mice, while 139 were down-regulated in both (fig. S4B). Of the DEGs in CM and NCM of HUPO mice, those up-regulated in both were enriched in the Gene Ontology (GO) pathway functions that may promote inflammation and adaptive immunity (fig. S4C), while no GO pathways exhibited a false discovery rate (FDR) P value < 1 for other combinations of genes up or down. In addition, supporting the role of adaptive immunity in the progression of arthritis, although conventional dendritic cells (cDC) were reduced in the spleens of HUPO mice (8), dendritic cells (DC), identified as CD45+CD11b+MHCII+CD64, were increased in the ST of HUPO mice with arthritis, compared to those without arthritis or controls (fig. S4D). Further, at 4 weeks before the onset of arthritis on clinical exam, cartilage proteoglycans were reduced in the cartilage of HUPO joints (fig. S4E), providing a potential antigenic source that may contribute to the anti-aggrecan antibodies previously observed in progressive HUPO arthritis (8).

Reduction of the HUPO FH1 subset is associated with arthritis

Next, we examined the role of nonhematopoietic cells in maintaining the STM niche in HUPO mice. Reciprocal BM chimeras were generated, and mice euthanized at 15 weeks after transplant demonstrated essentially complete repopulation of the ST with donor macrophages (fig. S5, A and B). As expected, the FH1 subset was markedly reduced in HUPO or control recipients that received HUPO BM (fig. S5C), while HUPO mice that received control BM exhibited a normal distribution of STM subsets. These observations confirm that the reduction of the FH1 subset is independent of mesenchymal cells in the synovial niche. Of interest, five of six mice receiving HUPO BM developed arthritis by 15 weeks (fig. S5D), while arthritis resolved after control BM transfer into the two HUPO mice that exhibited arthritis before transplant.

Control circulating monocytes fill the vacant FH1 niche in HUPO mice

To define the ability of wild-type monocytes to fill the HUPO tissue–resident macrophage niche, three types of parabionts were generated, linking HUPO-control, control-control, and HUPO-HUPO mice (Fig. 4A). Examination of long-lived B cells in the circulation and ankles suggested that CD45.2+ B cells had a slight survival advantage in the parabiont pairs (fig. S6A), as published (19). In the control-control parabionts, chimerism for circulating CM was 5 to 8% (fig. S6B), consistent with their short half-life, similar to previously described results (10, 20). The chimerism for NCM was higher, 33 to 35%, likely due to the longer half-life of NCM (12). For all STM subsets, chimerism was low in the control-control parabionts, although there was a slight survival advantage for the CD45.2+ donor cells in the FI2 and FH2 populations of the CD45.1 recipient mice but essentially no sharing from the partner parabiont in the FH1 tissue-resident populations (Fig. 4B).

Fig. 4 Circulating wild-type monocytes fill the vacant FH1 niche in HUPO mice and lead to amelioration of arthritis.

(A) Experimental design for parabiosis, linking the circulations of CD45.1+ and CD45.2+ mice. Photographer: Q.Q. Huang, NUFSM. (B) Percent of donor macrophages identified in the recipients of each STM subset from control-control parabionts. The total numbers of donor and recipient macrophages were identified at the bottom. (C) Representative histograms of the FH1 subsets from each of the donor-recipient pairs. (D) Percent of donor macrophages identified in recipient ankle STM subsets from the HUPO-control parabionts. (E) Mean fluorescence intensity (MFI) of MHCII expression in FH1 subset from HUPO mice comparing cells contributed from control donors and HUPO recipients. (F) Percent chimerism of macrophages identified in recipient ankle STM subsets from the HUPO-HUPO parabionts. (G and H) The individual joint inflammation scores at the time of surgery and 5 or 10 weeks after surgery for HUPO-control parabionts and HUPO-HUPO parabionts. Statistical analyses were performed by two-tailed Student’s t test (B, D, and F) and by paired t test (E, G, and H). Individual values and the means ± 1 SE are presented. *P < 0.05, **P < 0.01, and ***P < 0.001 among indicated groups.

Examination of the HUPO-control paired mice, euthanized at 5 and 10 weeks, revealed increased CD45.2+ HUPO CM in the control partner, likely due to increased HUPO CM, while the sharing of congenic NCM was 25 to 39% in both recipients (fig. S6B). An increased percentage of CD45.2+ HUPO-derived FI1 and FI2 macrophages was identified in the paired control parabionts, while the contribution of HUPO-derived monocytes progressively decreased from the FI3 (5%) through the FH1 subset (<1%) in the control recipients (Fig. 4C). In contrast, the contribution of control-derived macrophages in the HUPO mice increased from the FI3 (22%) through the FH1 subset (53%) (Fig. 4, C and D). Further, MHCII expression on the FH1 macrophages from the control donors was reduced compared to the expression on the recipient cells, consistent with a greater degree of differentiation of the donor cells (Fig. 4E). Although control monocytes contributed 53% of the cells to the HUPO FH1 subset, this represented only 19% of the total donor F4/80hi macrophages (calculated from Fig. 4D as FH1/FH1 + FH2), compared with 76% overall for control mice (from Fig. 2C), suggesting that cell-extrinsic factors may contribute to the reduced differentiation. One potential factor is that the apoptosis of CD45+ cells in chronically inflamed HUPO joints was reduced compared to controls (fig. S6C), which may contribute to reduced efferocytosis. The HUPO-HUPO parabionts demonstrated no significant differences between the pairs for any macrophage subset, and in contrast to the control-control parabionts, the HUPO-HUPO pairs demonstrated substantial sharing of the FH1 subsets (Fig. 4, B and E), which was significantly increased for the HUPO FHI pairs (fig. S6D). Arthritis was reduced in the HUPO parabionts paired with control mice, but not for the HUPO-HUPO pairs (Fig. 4, G and H). These observations demonstrate that HUPO monocytes readily entered the joints of control mice with very limited differentiation into F4/80hi tissue–resident macrophages. In contrast, wild-type monocytes, capable of expressing Flip, entered the joints of HUPO mice with arthritis, and differentiated into FH1 macrophages, acquiring a tissue-resident phenotype, which was associated with suppression of arthritis.

Synovial macrophage subsets in HUPO mice are more similar to CM

Next, we isolated and performed RNA-seq on circulating monocytes and STMs from HUPO mice with arthritis. On the basis of global gene expression, we observed that the HUPO FI2, FI3, FH2, and FH1 subsets appear more closely related to each other (Fig. 5A) than in controls (Fig. 1A). In addition, when comparing each of the HUPO subsets to their control counterparts, we observed that FI1 subsets, which maintain their already high monocyte similarity, were only marginally different between HUPO and control mice (Fig. 5B). In contrast, HUPO FH2 appears to be the most altered subset from its control counterpart, consistent with a high monocyte replacement in the HUPO FH2 population. In support of the monocyte origin of HUPO macrophages, the CM genes that were up-regulated in at least three of four HUPO macrophage subsets (n = 128) were enriched for interferon regulatory factor (IRF), PU.1-IRF, and CCAAT-enhancer-binding protein (C/EBP) binding motifs (fig. S7A).

Fig. 5 Transcriptional profiling supports the origin of HUPO STMs as circulating monocytes and identifies functional differences.

(A) PCA of gene expression for individual samples from HUPO mice. (B) Comparison of Pearson’s correlation coefficient of mean gene expression between each cell population from control and HUPO mice (n = 3). Total genes expressed (7778) were used for (A) and (B). (C) K-means clustering of DEGs (n = 2021) between control and HUPO subsets, identifying HUPO-signature (HS) genes. The genes from HUPO FI2 subset, which do not have a control counterpart, were included in the clustering. (D) Examples of genes important in macrophage efferocytosis and tissue residency, identified in (C), cluster V. (E) Examples of genes involved in the glycolytic pathway, identified in (C), cluster VI. Statistical analysis was performed by two-tailed Student’s t test (D and E). Individual values and the means ± 1 SE are presented. *P < 0.05, **P < 0.01, and ***P < 0.001 among indicated groups. n/a, no sample available.

Differential expression of genes between control and HUPO mice

To further determine how STMs differ between health and chronic inflammation, we defined 2021 DEGs, up- or down-regulated between HUPO and control mice in at least one of the myeloid subsets. We identified six HUPO-signature (HS) clusters demonstrating distinct expression patterns across STM subsets compared with control mice (Fig. 5C and data file S2). The HUPO FI2 subset, with no control counterpart, was included in the clustering. HS clusters I, IV, and V demonstrate expression patterns that were decreased in one or more HUPO subsets. Notably, cluster V genes, with decreased expression in HUPO FH2 and FH1 subsets, were consistent with functions associated with macrophage tissue residency during homeostasis (Fig. 5, C and D). In contrast, clusters II, III, and VI contain genes that generally were increased in expression in HUPO mice, which include GO pathways enriched for genes involved in inflammation, adenosine triphosphate metabolic processes, and glycolysis (Fig. 5E).

The tissue-resident macrophage phenotype is lost in HUPO and replaced by a proinflammatory profile

To further define the distinction between HUPO and control F4/80hi macrophages, we compared the genes expressed (4500) in either FH1 or FH2 subset in HUPO mice or controls (Fig. 6A). DEGs up-regulated in the HUPO FH1 subset alone included IL1b and neutrophil chemokine genes (Cxcl1, Cxcl2, and Cxcl3), which may contribute to the neutrophil recruitment into the HUPO joints (8). GO pathways increased in HUPO FH1 and FH2 macrophages were significantly enriched in genes involved in leukocyte migration and innate immune response (Fig. 6B). A significant number (173, P < 0.0001) of genes were down-regulated in both HUPO FH1 and FH2, including those important in maintaining macrophage tissue residency such as Csfr1, Cx3cr1, Timd4, and Vsig4. Further, proteins expressed by DEGs important in macrophage tissue residency were also reduced in the HUPO FH1 and/or FH2 subsets by flow cytometry (Fig. 6C). In addition, Mef2c was significantly reduced in HUPO FH2 macrophages, and DEGs down-regulated in both the HUPO FH1 and FH2 populations were enriched for myocyte enhancer factor 2 (MEF2) transcription factor (TF) binding motifs compared with those genes not down-regulated in either (fig. S7, B and C). These observations are in line with role of MEF2C in regulating the macrophage tissue–resident identity (21, 22). Notably, of the 356 genes from Fig. 1C, cluster IV that we previously associated with macrophage tissue–resident identity, 57 (P < 0.0001) were reduced in both HUPO FH1 and FH2 subsets (Fig. 6D), suggesting that the macrophage tissue residence phenotype was robustly diminished in HUPO mice.

Fig. 6 The tissue-resident macrophage phenotype is lost in F4/80hi HUPO STMs and in RA ST.

(A) Scatterplot showing fold change of genes expressed in HUPO/control FH1/FH2 subsets. DEGs and their numbers are in colors, and select genes are labeled. (B) Top GO terms for genes in red and dark blue from (A). (C) Flow cytometry for protein expression of genes down-regulated in HUPO mice from (A), as MFI and % positive (n = 4 to 14 and means ± 1 SE). (D) Overlap of 356 genes from the control STM cluster IV in Fig. 1C with genes in (A). (E) Following (16 to 40 hours) ankle injection of Dil-lip, control, and HUPO cells were harvested and MHCII on the FH2 subsets determined for Dil+ and Dil− cells (representative and individual plus means ± 1 SE). *P < 0.05, **P < 0.01, and ***P < 0.001 among indicated groups, for C and E. (F and G) Module scores (box plots) of top 20 up-regulated (F) or 20 down-regulated genes (G) in HUPO FH1 and FH2 (A) and the representative genes (violin plots) expressed in osteoarthritis (OA) and RA leukocyte–poor or RA leukocyte–rich ST (23). Statistics: Two-tailed Student’s t test (C and E); comparison with expected background distribution from 10,000 permutations (A and D) and pairwise Wilcoxon rank (F and G). Absolute P values between groups (F and G).

To identify a mechanism for the reduction of the FH1 subset in HUPO mice, we examined the ability of phagocytosis of liposomes, as a surrogate for phagocytosis of apoptotic cells, to promote the differentiation of FH2 macrophages, determined by the reduction of MHCII expression. This approach was chosen since the expression of genes for receptors contributing to efferocytosis was reduced in HUPO FH2 macrophages (Fig. 5D). Dil-lip injected into the ankles of control mice resulted in the reduction of MHCII expression on FH2 macrophages that phagocytosed Dil-lip compared to those that did not. No reduction of MHCII was observed on HUPO FH2 macrophages that phagocytosed Dil-lip, and this was significantly different (P < 0.001) compared with the controls (Fig. 6E). These observations directly demonstrate that phagocytosis of liposomes was capable of promoting macrophage differentiation under homeostatic conditions, which was not observed with HUPO FH2 macrophages.

To determine whether the differences observed in FHI and FH2 macrophages from HUPO-associated arthritis were conserved in human disease, we compared gene expression in single cell (sc) RNA-seq previously performed on RA ST (23). Overall, we found that of the 20 genes that were most up- or down-regulated in HUPO F4/80hi macrophages, the module scores of those up-regulated were modestly increased, while those down-regulated were significantly reduced, in monocytes from leukocyte-rich RA ST (fig. S8 and Fig 6, F and G). Specifically, we noted that the expression of genes associated with increased HUPO expression, including S100A9 and VEGFA, was higher in ST monocytes from patients with leukocyte-rich RA ST (Fig. 6F). On the other hand, genes associated with decreased HUPO expression, such as CX3CR1, TIMD4, VSIG4, and TGFBR2, were significantly decreased in expression in leukocyte-rich RA ST, suggesting that the loss of tissue residence phenotype is conserved in highly inflammatory human disease (Fig. 6F). Together, these observations suggest that the reduction of the FH1 subset in HUPO mice emptied a niche, into which monocyte-derived FH2 macrophages entered but, in the absence of Flip and under chronic inflammatory conditions, demonstrated limited ability to differentiate into bona fide tissue-resident macrophages, capable of suppressing chronic inflammation, which may be relevant to the pathogenesis of RA.


Our observations suggest a novel mechanism for the transition from preclinical disease to active RA. During homeostasis, the FH1 subset was the dominant tissue-resident population, while the smaller FH2 subset was the F4/80hi population that was slowly proliferating. In HUPO mice, macrophages are necessary for the initiation of disease, while T and/or B lymphocytes are required for progression (8). Here, we demonstrate that the HUPO FH1 subset was reduced before and after disease onset, while the other subsets were variably increased shortly after arthritis onset and during chronic inflammation. BrdU labeling, transcriptional profiling, and parabiosis experiments established that circulating monocytes readily differentiated into F4/80hi macrophages in HUPO mice, which was minimal during homeostasis, but with the reduction of Flip, in the environment of chronic inflammation, exhibited limited ability to differentiate into bona fide FH1 tissue–resident macrophages.

Recent studies demonstrate that, like most tissue-resident populations, except microglia, which renew from fetal hematopoietic stem cells, CX3CR1+ synovial lining macrophages derive during early embryonic development (9, 24). Our analysis demonstrates that the FH1 and FH2 populations represent long-lived tissue-resident macrophages, which maintain their population in steady state via local proliferation of the FH2 subset, with minimal replenishment from circulating monocytes. These findings support recent observations that identify CX3CR1+ lining and CX3CR1 interstitial STMs, consistent with our CX3CR1+ FH1 and CX3CR1lo FH2 subsets, as the populations maintained with limited contribution from circulating monocytes (24). Further, supporting the similarity of the interstitial CX3CR1 and FH2 populations, each was the primary subset that proliferated under homeostatic conditions. MHCII expression on most F4/80hi populations of tissue-resident macrophages is relatively homogeneous, except for dermal macrophages, which demonstrate low and high MHCII subsets (9), similar to the STMs under steady-state conditions. Previous reports have shown that phagocytosis of apoptotic cells facilitates the differentiation of the MHCII+ subset to become MHCII and is important in maintaining tissue residency and tissue homeostasis (2527). The differentiation of CX3CR1MHCII+ macrophages into CX3CR1+ synovial lining macrophages (24) is consistent with the transition of FH2 into FH1 macrophages under homeostatic conditions. We speculate that the CD45+ cell apoptosis noted in the joints of the wild-type mice may contribute to this process through efferocytosis (27). Our observations suggest the ST-resident macrophages function to suppress chronic inflammation, not only serving as a barrier and dampening serum transfer–induced arthritis (24) but also capable of suppressing chronic inflammation following recruitment and differentiation of Flip-replete monocytes to the ST.

In contrast to homeostatic conditions, the BrdU and parabiosis experiments demonstrated that HUPO circulating monocytes entered the joints readily differentiating into F4/80hi macrophages; however, the transition from FH2 to fully functional FH1 cells was impaired. The lack of enhanced local proliferation of the F4/80hi HUPO macrophages 30 min after BrdU administration is distinct from observations in acute serum transfer–induced arthritis (24) and the early inflammatory phase of atherosclerosis in which monocytes are recruited, differentiate into macrophages, and expand by proliferation (28). Further, bacterial infection results in Kupffer cell necroptosis and recruited monocytes proliferate, differentiating into tissue-resident macrophages (29). In contrast, in HUPO arthritis, which is chronic, the influx of monocytes primarily accounted for the expansion of the FH2 population. The reduction of FI3 and FH2 HUPO macrophages, concurrent with the reduction of NCM, but not CM, following the injection of clodronate liposomes, suggests that these macrophage subsets derive from NCM, although this was not directly documented. The HUPO monocytes entering the joints were highly enriched in pathways of innate immunity, antigen presentation, and interferon signaling, which, together with the increase of DCs in the HUPO ST and the systemic reduction of regulatory T cells present in HUPO mice, contribute to the development of autoantibodies to joint constituents such as aggrecan and the progression of arthritis (8). This scenario is distinct from the setting where the niche is disrupted under noninflammatory conditions, into which homeostatic monocytes enter, acquire a tissue resident phenotype, and restore the niche (30, 31).

In HUPO mice with arthritis, the FH1 population was greatly reduced, while the FH2 population was expanded. Although Flip was reduced in both FH1 and FH2 macrophages under inflammatory conditions, neither population exhibited increased apoptosis compared with control mice. Previously, we demonstrated that apoptosis of human monocytes was rescued by the increase of FLIP observed during in vitro differentiation into macrophages (15). The reduction of FLIP in human in vitro–differentiated macrophages results in Fas-mediated apoptosis. In contrast, using lineage negative murine BM progenitors, we demonstrated that the reduction of Flip prevented macrophage differentiation in vitro, which was not associated with increased apoptosis and was not rescued by caspase inhibitors (6). Together, these observations support the potential role of Flip in monocyte to tissue-resident macrophage differentiation, under chronic inflammatory conditions, which promoted macrophage survival despite the reduction of Flip.

In HUPO mice, the FH1 subset replenished to a limited degree by circulating monocytes, determined by BrdU and parabiosis, and demonstrated a restricted tissue-resident transcriptional profile. A potential mechanism preventing HUPO F4/80hi macrophages from attaining tissue residency may be the reduction of the requisite genes. For example, CD115/CSFR1 was reduced in HUPO F4/80hi macrophages, together with TFs downstream of CD115, such as Klf2 and Klf4, which are involved in the development of macrophage tissue residency and the ability to silently clear apoptotic cells (27). Efferocytosis is known to drive the differentiation of monocytes to tissue-resident macrophages under homeostatic conditions, in an acute self-limited model of arthritis and following myocardial ischemia (2527). Our data suggest that decreased phagocytosis of apoptotic cells may also contribute to the decreased FH1 population in HUPO mice. Receptors responsible for efferocytosis including CD206 and CD163, which are highly expressed on phagocytic macrophages, were reduced in HUPO FH2 macrophages compared with controls, and apoptotic CD45+ cells were greatly reduced in the chronic inflammation of the HUPO joints. The forced reduction of CD206 reduces efferocytosis (25), and mice deficient in CD206 develop more severe experimental arthritis (32). CD163, a scavenger receptor, is important in suppressing inflammation (33). Further, we directly documented the inability of phagocytosis of liposomes to drive the reduction of MHCII on HUPO FH2 macrophages, and this decline is a phenotypic characteristic of ST-resident macrophage differentiation observed with wild-type FH2 macrophages (26), although phagocytosis of these commercially available lysosomes is not known to be mediated by a specific receptor. Our observations, combined with the parabiosis data, demonstrate that, although the molecular mechanism remains to be defined, Flip was necessary, but not sufficient, for the differentiating monocytes to fully acquire a tissue-resident, homeostatic phenotype during chronic inflammation.

Our observations concerning the role of STMs and circulating monocytes in the initiation and progression of arthritis in HUPO mice appear relevant to the pathogenesis of RA, although FLIP is highly expressed in RA STMs (5). The mechanisms targeting ST in the transition from preclinical to clinical disease are not known. Our data suggest a novel mechanism, which reduction of tissue-resident macrophages predisposes joints to the development of arthritis, by opening a niche permitting the influx of activated monocytes and other inflammatory cells. Consistent with this notion, the FH1 subset was reduced before the onset of clinical arthritis and restoration of this subset from wild-type mice suppressed inflammation in HUPO mice. HUPO monocytes expressed a transcriptional profile enriched in pathways that support inflammation and adaptive immunity, which likely contributed to the progression of disease. Similarly, monocytes from patients with RA are enriched in pathways involved in interferon signaling, inflammatory response, and anti-apoptosis (34), supporting a role for activated monocytes in the pathogenesis of RA. Before the onset of RA, patients also exhibit circulating autoantibodies and inflammatory mediators (1, 2). Further, the incidence of RA increases with age (35), while murine tissue–resident alveolar and human BM CD68+ macrophages (36, 37) and mouse STMs all decrease with age. In addition, MHCII tissue–resident–like macrophages are reduced in patients with active RA compared with those with osteoarthritis (24), and genes relevant to macrophage tissue residency such as CX3CR1, VSIG4, TIMD4, and TGFBR2 are reduced in leukocyte-rich RA ST. Also, circulating monocytes from patients with RA exhibit a defect in their ability to differentiate into M2-like macrophages, mediated by miR-155, resulting in reduced expression of CD206 and CD163, and increased proinflammatory mediators (38). Consistent with our interpretation, recently published data demonstrate that when therapy is stopped or reduced in individuals with RA while their disease is in remission, those with increased MerTK+CD206+CD163+ tissue–resident macrophages, expressing transcriptomes enriched in anti-inflammatory signatures, on ST biopsy, are less likely to experience a recurrence of the joint inflammation (39). Together, these observations identify a critical role for tissue-resident macrophages in the pathogenesis of chronic arthritis and raise the possibility for targeted treatments that prevent disease from developing in individuals with preclinical RA or maintaining remission by modulation of the STM populations.



We generated the CD11c-Flip-KO (referred to as HUPO) mouse line in C57BL/6 background, genotyping as Flipflox/floxCD11ccre (8). HUPO mice are generated by crossing Flipflox/flox mice with CD11c-Cre-GFP transgenic mice (CD11ccre) [C57BL/6J-Tg (Itgax-cre,-EGFP) 4097Ach/J, Jackson stock 007567] (6). Control mice were littermates or age/gender-matched mice that genotyped as Flipflox/+CD11ccre or Flipflox/floxCD11c+. HUPO mice age ≥ 22 and ≤ 37 (median = 31) weeks old were used unless otherwise stated. For some experiments, HUPO mice (4 to 16 weeks) with no arthritis or within 1 to 2 or 4 to 6 weeks of arthritis onset were used. HUPO mice developed spontaneous arthritis beginning at 6 weeks of age, identified as swollen and deformed joints. Arthritis frequency and severity increased through 5 months, with no difference in incidence or severity between females and males that were both used for these studies.

HUPO or control mice on C57BL/6 CD45.1+ background were generated by crossing with CD45.1 congenic strain (B6.SJL-Ptprca Pepcb/BoyJ, Jackson stock 002014). All mice were breed on the C57BL/6 background. All genotyping was performed by polymerase chain reaction using genomic DNA extracted from tail biopsies. CD45.1 or CD45.2 background was determined by flow cytometry. The mice were bred and maintained in the Northwestern University barrier animal facility, and all procedures followed ethical guidelines and approved by Northwestern IACUC.

Clinical evaluation of arthritis

The incidence and severity of HUPO arthritis were assessed by clinical examination (8, 40). The clinical score was quantitated from the sum of joint swelling/inflammation (graded 0 to 3 per each limb) and joint deformity (including toe flexion, contraction, and shortening; 0 to 3 per each limb) and the grip strength (0 to 4), the maximum score being 28 (8). Incidence was defined as at least one swollen, inflamed, or deformed joint.

Histologic analysis

Mouse interphalangeal joints of the hind ankles were used for histologic analysis. Ankles were fixed by 10% formalin and decalcified in 7% EDTA/10% formalin (w/v) for 2 weeks. After procession and paraffin embedding, sections were analyzed by hematoxylin and eosin (H&E) and immunohistochemistry of Safranin-O staining, performed by Northwestern Mouse Histology and Phenotyping Laboratory. The H&E slides were scored on a scale of 0 to 5 for joint and extraarticular inflammation and cartilage destruction, as previously described (8). Macrophages in ankles were identified by immunohistochemistry using anti-F4/80 (1:250; Cell Signaling, cat. no. 70076) or an equal amount of isotype-matched control immunoglobulin G (Abcam, cat. no. ab37415) antibodies. F4/80-positive mononuclear cells were quantified by the average of three views of ankle synovial linings or sublinings areas at ×40 magnification, calculated per 0.01 mm2. Ankles used for Safranin-O staining were decalcified with formic acid for 1 day, before processing and paraffin embedding. The loss of proteoglycan in cartilage was scored on a 0 to 4 scale, with no loss 0, + mild (<25%), 2+ moderate (25 to 50%), 3+ severe (50 to 75%), and 4+ (>75%). All histology was scored by a pathologist blinded to the experimental design. Images were acquired by a Nikon microscope (Nikon Inc.), with SPOT5.2 software (SPOT Imaging, Sterling Heights, Michigan).

Flow cytometric immunophenotyping

Circulating monocytes and the BM precursors, as well as synovial macrophages, were characterized by flow cytometry using multicolor fluorochrome-conjugated antibodies to cell surface and intracellular markers. Blood was drawn by cardiac puncture immediately after euthanizing or submandibular puncture after isoflurane anesthesia. EDTA-anticoagulated whole blood was used for flow. BM cells were isolated from femurs after lysing the red blood cells. Synovial cells were dissected from ankle joints with the methods modified from our earlier publication (26). Ankles were cut 3 mm above the heel, and skin was removed from the feet. The toes were disarticulated by pulling with blunt forceps, and tibiotalar joint was opened via posterior access to expose the synovial lining. The opened BM cavity in the tibia was thoroughly flushed with Hanks’ balanced salt solution to remove BM cells. The dissected joint was incubated in collagenase D (1 mg/ml) for 60 min at 37°C. The released cells were filtered through a 40-mm nylon mesh, and the resulting single-cell suspensions were used.

BM monocyte progenitors and macrophage dendritic cell precursors and BMMs were defined using cocktails containing antibodies to CD117, CD115, CD135, CD11b, and Ly6C gating on the Lin population (18). Circulating monocytes were defined using cocktails containing antibodies to CD45, CD11b, CD115, Ly6G, Ly6C, CD62L, F4/80, Cx3cr1, and MHCII. Synovial macrophages were defined using cocktails containing CD45, CD11b, Ly6G, Ly6C, MHC class II (I-A/I-E, MHCII), F4/80, CD64, Siglec F, CD11c, and CX3CR1. Alveolar macrophages were defined as Siglec F+CD64+F4/80+CD11c+CD11b. Synovial macrophage CD115, CD206, CD163, and TGFBR2 were identified by intracellular staining (41). The live/dead cell marker Aqua (Invitrogen) accompanied every run. Data were acquired on BD LSR II flow cytometer (BD Biosciences), and analysis was performed using FlowJo software (Tree Star Inc.).

In vivo proliferation and apoptosis

For BrdU (Invitrogen) incorporation assays, two injections of BrdU [1 mg/250 μl of phosphate-buffered saline (PBS)] were administered intraperitoneally on day 0. BM, blood, and ankle cells were harvested at days 1, 3, and 5 after BrdU administration. BrdU incorporation was determined by flow cytometry using the BrdU Flow Kits (BD Biosciences) and cell surface markers. In addition, BrdU was administrated once intravenously, and the cells were harvested 30 min afterward. Synovial macrophage apoptosis was assessed with annexin V and cell survival by the exclusion of 7-aminoactinomycin (BD Pharmingen).

In vivo monocyte and synovial macrophage depletion

Littermate control or HUPO mice with established arthritis were administered clodronate-loaded liposomes (Liposoma B.V.) by retro-orbital injection at 200 μl or an equal volume of PBS. Blood and ankle cells were harvested at days 1, 2, and 3 and were characterized by flow cytometry.

In vivo fluorescent Dil-lip uptake

Littermate control or HUPO mice with established arthritis were administered fluorescent Dil-lip (Liposoma B.V.) by retro-orbital injection at 200 μl per mouse or intra-articular injection at 20 μl per ankle, two ankles per mouse. Blood and ankle cells were harvested at indicated time points after injection and characterized by flow cytometry.

BM reconstitution

BM chimeras were established in a CD45.1/CD45.2 mismatched manner. BM recipients, control or HUPO on CD45.1 background were lethally irradiated (γ-radiation, 1100 rads). After 4 hours, 5 × 106 donor whole BM cells from CD45.2 HUPO or control mice were administered by retro-orbital injection. Recipients received sulfamethoxazole (50 mg/ml) and trimethoprim (8 mg/ml) in the drinking water for 8 weeks. Arthritis was evaluated by clinical scoring. Blood and ankles were harvested at 15 weeks and characterized by flow cytometry.


CD45.1/CD45.2 mismatched mice, control-control (three pairs), HUPO-control (six pairs), and HUPO-HUPO (three pairs) of the same gender and age and similar in body weight were paired by parabiosis (42) performed by the Microsurgery Core of Northwestern University. Food and water were supplied at the bottom of cages, and 1-ml saline per mouse was subcutaneously injected as needed between 1 and 4 weeks after surgery, depending on the body weight. Two weeks after procedures, chimerism was determined by flow cytometry of peripheral blood cells. Arthritis was evaluated by clinical scoring starting 2 weeks after the procedure. Blood and ankles were harvested 5 or 10 weeks after parabiosis, and the cell populations were analyzed by flow cytometry.

Preparation of the RNA-seq library

HUPO mice with established arthritis and littermate or age/gender-matched control mice were euthanized to harvest blood and ankle cells for RNA-seq. CM and NCM and five subsets of STMs from ankles were stained by multicolor fluorochrome-conjugated antibodies as described and then sorted at the Northwestern University Flow Cytometry Core Facility using a FACSAria III instrument (BD Biosciences). Each HUPO sample was from an individual mouse; however, some ankle samples from control mice were combined because of the limited number of cells. RNA was extracted from each sorted cell population (>400 cell count) using the Arcturus PicoPure RNA Isolation Kit (Applied Biosciences) according to the manufacturer’s instructions. Total RNA was used for library construction, which included 49 samples across six control and seven HUPO populations of cells. Full-length cDNA synthesis and amplification were carried out with the Clontech SMART-Seq v4 Ultra Low Input RNA Kit. Subsequently, Illumina sequencing libraries were prepared from the amplified full-length cDNA with the Nextera XT DNA Library Preparation Kit. Before sequencing, the prepared libraries were quantified with Qubit and validated on a Bioanalyzer with a high sensitivity DNA chip. The sequencing of the libraries used an Illumina NextSeq 500 NGS System. Single 75-bp reads were generated with dual indexing, and the libraries were sequenced to an average depth of 21.8 million reads. These procedures were performed at the NUSeq core Facility of Northwestern University.

RNA-seq analysis

The sequencing library was demultiplexed, and the quality of DNA reads was evaluated using FastQC. Adapters were trimmed, and reads of poor quality or aligning to ribosomal RNA sequences were filtered. The sequenced reads were aligned to the Mus musculus genome (mm10) using STAR (43). Read counts for each gene were calculated using htseq-count (44) in conjunction with a gene annotation file for 23,337 genes obtained from UCSC (University of California Santa Cruz; Raw gene expression counts were normalized to fragments per kilobase per million (FPKM) using cufflinks (45). These procedures were performed at the Quantitative Data Science Core of Northwestern University.

Quality control for the 49 samples was performed excluding five low-quality samples (alignment < 82%, duplicates > 66%, and mapped reads < 2.3 × 106) from groups with >3 replicates. To maintain consistent numbers of replicates (n = 3) in each experimental group, we removed the least correlated sample from an additional three control mice under the assumption that the most similar samples better reflect steady-state conditions. We also removed two HUPO samples from F4/80hi subsets that appeared to have been contaminated on the basis of the high expression of neutrophil-specific genes. Thus, 39 samples reflecting three replicates for each of the 13 groups remained. We defined 7778 “expressed genes” with log2 (FPKM + 1) ≥ 4 in at least 2 of the 39 individual samples. Mixed sexes were used in the experiments, since the HUPO phenotype was not significantly different between male and female mice. Further, removing genes on the sex chromosomes (X and Y) did not affect the results of the transcriptional analysis.

K-means clustering for control mice was performed for genes that fulfilled the following criteria: (i) mean expression across replicates in at least one of the subsets was [log2 (FPKM + 1)] ≥ 4; (ii) there was a log2 fold change of ≥1 in one subset compared with any other; and (iii) P < 0.05 by analysis of variance (ANOVA) across subsets. DEGs for a given population of myeloid cells between HUPO and control mice were defined by the following criteria: (i) The mean expression for a given population in either group was ≥4; (ii) the magnitude of the log2 fold change between HUPO and control was ≥1; and (iii) P < 0.05 by t test between HUPO and control in any one of the five subsets. Because of the low cell numbers, there was no FI2 subset harvested from control mice for RNA-seq analysis. Therefore, the expression of the HUPO FI2 subset is given without a matching control.

The GENE-E software ( was used for the pairwise Pearson’s correlation and K-means clustering analyses, performed using the default settings. PCA was performed using the prcomp function in R with the FPKM matrix of the expressed genes. GO associations were determined by Gorilla (

To identify TF binding motifs, we used the function with default parameters in the HOMER software package. To identify the potential contributions of TF regulation to HUPO macrophages, we compared the proportion of genes with selected TF binding motifs between the up- or down-regulated genes and those not significantly changed in HUPO macrophage subsets.

Human scRNA-seq analysis

A processed human scRNA-seq dataset was obtained from published data (23). The top 20 up- and down-regulated genes ranked by the sum of fold changes in the HUPO FH1 and FH2 subset from Fig. 6A, which have orthologs present in the human dataset (23), were used for module score calculation performed using FindModuleScore function in Seurat v3.1.0 package with default parameters.

Statistical analysis

All quantitative data are presented as means ± SEM. Statistical analysis between two groups was performed with unpaired two-tailed Student’s t test. For multiple comparisons, one-way ANOVA was used followed by Tukey’s pairwise mean comparison. Correlations were determined by Pearson’s linear correlation. The Bonferroni correction was performed when a single value was compared with multiple variables, and the corrected P value (pc) was presented. To investigate the enrichment of gene sets of interest, the significance of the observed gene numbers was determined by performing permutations (10,000×) to generate putative background distributions. Comparison of module scores and single-cell gene expressions between disease groups was performed with pairwise Wilcoxon rank-sum test. All significance levels were set at P < 0.05.


Supplementary material for this article is available at

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank D. Fang for review of this manuscript, M. Schipma and the NUSeq Core for assistance, and P. M. Newberg for generous support. Funding: This work was supported by NIH R01AR070025 and AR048269 and the Solovy/Arthritis Research Society Endowment to R.M.P.; grants from the Arthritis National Research Foundation, American Lung Association (514724), Scleroderma Foundation, American Federation for Aging, American Heart Association (18CDA34110224), and an ATS Foundation/Mallinckrodt Pharmaceuticals Research Fellowship to D.R.W.; and NIH grants HL135124, AG049665, AI135964 and Department of Defense grant PR141319 to A.V.M. Author contributions: R.M.P. and Q.-Q.H. conceived and designed research studies. Q.-Q.H. performed major data analysis. D.R.W. supervised the RNA-seq and transcriptional analysis. Q.-Q.H. and R.D. performed the experiments and data acquisition. S.-Y.C. and Q.S. contributed to RNA-seq analysis. Q.M. performed histologic analysis. A.V.M. provided essential advice and assisted with data analysis. R.M.P., D.R.W., and Q.-Q.H. interpreted data and wrote the manuscript. R.M.P. supervised this study. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials or archived in the GEO database, accession number GSE158446. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article