Macrophage (MØ) biology is routinely modelled in the peritoneal cavity, a vascular tissue readily infiltrated by leukocytes during inflammation. After several decades of study, no consensus has emerged regarding the importance of in situ proliferation versus peripheral monocyte recruitment for the maintenance of tissue resident MØs. By applying specific measures of mitosis, we have monitored tissue MØ proliferation during newborn development, adulthood and acute resolving inflammation in young adult mice. Despite the vascular nature of the tissue and ease of peripheral leukocyte entry, tissue MØs in the newborn increase in number by local proliferation. On the contrary, in the adult, tissue MØ proliferation is considerably reduced and most likely provides homeostatic control of cell numbers. Importantly, during an acute inflammatory response, when substantial numbers of inflammatory MØs are recruited from the circulation, tissue-resident MØs survive and then undergo a transient and intense proliferative burst in situ to repopulate the tissue. Our data indicate that local proliferation is a general mechanism for the self-sufficient renewal of tissue MØs during development and acute inflammation and not one restricted to non-vascular tissues, which has implications for the therapeutic modulation of MØ activity during the resolution of inflammation.
The peritoneal cavity, a vascular environment extensively used for the study of macrophage (MØ) biology, receives a large influx of inflammatory cells (including monocytes/MØs) during inflammation, such as that induced with zymosan. Numbers of recoverable tissue-resident (Res) MØs drop during the first hours of inflammation (‘the disappearance reaction’) 1–3. ResMØ-like cells become recoverable again about 2 days after the induction of inflammation in an environment mixed with inflammatory peripherally recruited MØs 2, 4. The MØ disappearance reaction has been attributed to several factors: increased tissue adherence, emigration to draining lymph nodes and/or cell death 1. Historically, many attempts have been made to document the origins of peritoneal MØs under steady-state and inflammatory conditions. However, these have produced conflicting results often arguing for negligible local proliferation and the maintenance of ResMØs from the periphery 5, 6. Complicating these studies, subsets of peritoneal ResMØs are binucleate or can occasionally be found in the act of phagocytosis of apoptotic and other debris, casting doubt on the apparent presence of DNA contents consistent with active cell cycle.
The generally accepted model is that ResMØs are renewed in tissues by myeloid restricted BM hematopoietic progenitors 7–9. Some evidence has been presented for local proliferation of ResMØs in a variety of tissues, but the results were often controversial and varied with models used 8. More recently, it has been proposed that local proliferation may be more important in tissues where leukocyte traffic is restricted 10 and this has been demonstrated in the epidermis 10 and brain 11, 12. Very recently, proliferation of ResMØs has been observed during parasite infection 13.
In this study, applying specific measures of proliferation to avoid artifacts, we have determined the mechanism of peritoneal ResMØ renewal during an episode of acute-resolving inflammation. We demonstrate that in the presence of significant numbers of ‘inflammation-associated’ (Inf) MØs, ResMØs survive the inflammatory response and then experience a transient quantifiable proliferative burst within the tissue that rapidly repopulates it with ResMØs. The data indicate a negligible contribution of either peripherally recruited MØs or an alternative source to the pool of ResMØs. Furthermore, the studies in newborn mice are consistent with local proliferation as a source of tissue MØs, as seen with the Langerhans cells of the skin 10. These data argue that the self-sustenance by local proliferation of mature tissue MØs may represent a general paradigm and may not simply be restricted to tissue that has limited or controlled leukocyte access.
Presence of additional non-cell cycle-associated DNA content in ResMØs
The examination of mouse peritoneal ResMØs on cytospin preparations highlighted a small number that were binucleate, whereas an even smaller number were multinucleate (Fig. 1A and data not shown). Very rare mitotic events were also seen (Fig. 1A). Additionally, some MØs could be visualized in the process of phagocytic clearance of apoptotic cells or other ill-defined matter (data not shown).
Gating on ResMØs by their F4/80highCD11bhigh phenotype with great care to exclude doublets, eosinophils 3 (as explained in Definition of ResMØs in complex inflammatory environments), and DCs 4, 14, we examined Ki67 expression, which is restricted to the active phases of the cell cycle (G1, S, G2 and M) 15. Ki67 staining revealed Ki67− MØs with 4N DNA content, assumed to be ‘binucleate’ MØs, and cells with DNA content between 2N and 4N, which were tentatively termed ‘phagocytic’ (Fig. 1B). Since Ki67 staining was clear, but had relatively poor resolution, we examined histone H3, which is selectively phosphorylated histone H3 (pHH3) at a discrete stage of mitosis 16. Adult mice had only a low proportion (<0.03%) of pHH3+ ResMØs (Fig. 1B). Discrete pHH3 staining was associated with very high levels of Ki67 antigen detection (Fig. 1B). Quantification of cell cycle and basal (>2N) DNA content of adult mice (Fig. 1C) identified a greater number of Ki67− cells with >2N DNA staining than the number of cells in S, G2 or M phases of cell cycle, but overall the data demonstrate a clear low-level proliferation of ResMØs in naïve adult tissue. When quantified, 4.62±1.04% (mean±SEM) were in G1 (2N DNA content and Ki67+); 0.19±0.07% (mean±SEM) were in S phase (between 2N and 4N DNA content and Ki67+); and 0.20±0.07% (mean±SEM) were in G2/M (4N DNA content and Ki67+).
Definition of ResMØs in complex inflammatory environments
The conclusion that the F4/80highCD11bhigh phenotype selectively represents ResMØs in the peritoneal cavity 2–4, 17 was tested in both naïve and immune-challenged mice by using the additional ResMØ-selectively expressed marker Tim4 18. Tim4 is expressed on the vast majority (>95%) of ResMØs in the peritoneal cavity of naïve adult mice (Fig. 2). Care was taken to exclude doublets, as detailed in Materials and methods and as shown in Fig. 2A. Having confirmed the identifications, the distinctive FSC/SSC profile of eosinophils was used to remove them from subsequent MØ analysis (Fig. 2B). The use of the Tim4 marker validated the F4/80high phenotype of ResMØs, and the F4/80low characteristic of ResDCs 4, 14, 17 in naïve mice (Fig. 2C). Furthermore, Tim4 remained restricted to F4/80high ResMØs during inflammation when a large number of F4/80+ InfMØs were present (Fig. 2C).
Proliferation of ResMØs during ontogeny
We next used pHH3 to screen for alterations in peritoneal ResMØ proliferation by flow cytometry. First, we quantified the number and proportion of F4/80high ResMØs in mice from newborn (P14-15, 2 wk) to mature adulthood (12–16 wk old) (Fig. 3A). An increase in the number of ResMØs was evident between 2 and 4 wk of age, a time when the ResMØs were the dominant leukocyte population in the tissue. As the mice aged, the proportion of ResMØs decreased but with little change in overall MØ numbers (Fig. 3A).
While the analysis of DNA content indicated a threefold increase in the proportion of ResMØs with >2N in the 2-wk-old mice (5.15±1.65%; mean±SEM), compared with the 12- to 16-wk-old mice (1.16±0.12%; mean±SEM) (Fig. 3B), the number of pHH3+ ResMØs increased 20-fold (2 versus 12–16 wk mice: 0.25±0.07 versus 0.011±0.003% respectively; mean±SEM) (Fig. 3B), indicating the superior resolution of this method for scoring proliferation. Interestingly, when the proportion of ResMØs with DNA content >2N was correlated with those that were pHH3+, clear linear correlations existed with cells from the younger mice that were not readily detected with cells from the older mice (Fig. 3C).
Proliferative burst of ResMØs after acute inflammation
We next examined the impact of an acute inflammatory insult (i.p. zymosan) on the proliferation of ResMØs. We examined young adult mice to maximize the usefulness of DNA content analysis alongside pHH3 staining (Fig. 4). In the early phases of zymosan peritonitis (e.g. 18 h post-injection), recoverable InfMØs greatly outnumbered the ResMØs (>5-fold) (data not shown). Most neutrophils were cleared by 48 h (data not shown). By 48 and 72 h, InfMØs numbers were declining and ResMØ numbers returned to normal levels (Fig. 4A). We analyzed DNA content and pHH3 staining in the ResMØ and InfMØ gates of all the mice (ResDCs were present in the InfMØ gate of naïve mice). Representative flow cytometric plots indicated a temporally restricted increase in DNA content concurrent with a substantial increase in pHH3 staining in ResMØs. This was noticeably absent in InfMØs (Fig. 4B). Analysis of cell cycle by the DNA content of ResMØs at 48 h post-injection showed significant evidence of in situ proliferation (Fig. 4C). Two-way ANOVA analysis of the proportion of both ResMØs and InfMØs with DNA content >2N or that were pHH3+ confirmed the presence of a marked ResMØ-specific proliferative burst, immediately after the clearance of neutrophils and corresponding with repopulation of the tissue with normal ResMØ numbers (Fig. 4D). Forty-eight hours after zymosan, the proportion of ResMØs with >2N DNA content was 13.85±1.62% (mean±SEM) and the proportion that were pHH3+with a 4N DNA content was 0.64±0.07% (mean±SEM). Analysis of Ki67 expression showed that 41.44±6.99% (mean±SEM) of ResMØs were Ki67+ (n=3; from one of the two similar experiments). In similar naïve young adults, 3.83±0.44% (mean±SEM) of ResMØs were Ki67+ (n=10). In all cases, the proportion of cells with DNA content >2N correlated well with pHH3 staining (Fig. 4E).
A significant proportion of ResMØs are generated by proliferation during inflammation
The proportion of pHH3+ ResMØs during resolution of inflammation (0.64±0.07%, mean±SEM) was approximately a third of that observed in a fast growing immortalized cell line (1.64±0.17% of MØs P-S6 are pHH3+ under immortalized growth conditions, mean±SEM; n=5). To better quantify the amount of proliferation, we included BrdU in the drinking water from the time of zymosan challenge. Analysis of BrdU incorporation after 72 h showed significant uptake by ResMØs (∼50% of cells), but a much more modest uptake by InfMØs (Fig. 5A and B).
Mature ResMØs in the naïve tissue survive inflammation and proliferate to reconstitute the tissue
When adoptively transferred from one mouse to another, ResMØs are known to be long-lived. We confirmed this by showing that half of the transferred cells (detected by CD45 allotypes) were detectable in hosts 5 days after transfer (Fig. 6A). We adoptively transferred cells from naïve CD45.1+ congenic 129S6/SvEv mice into naïve wild type (CD45.2+) 129S6/SvEv mice. One day later, the mice were injected with zymosan or left unchallenged. Mice with zymosan were sacrificed after 4 or 48 h of zymosan-induced peritonitis (Fig. 6B). Non-challenged mice were also sacrificed at 4 h to monitor the baseline persistence of donor ResMØs and the MØ disappearance reaction. After the induction of peritonitis (48 h), donor ResMØs were recoverable from the peritoneal cavity, whereas InfMØs were almost exclusively of host origin (Fig. 6C).
Quantification of ResMØ numbers in these adoptive transfer experiments showed that donor ResMØs behaved exactly as the host ResMØs, with approximately 80% disappearing 4 h after zymosan, but repopulating the tissue 48 h after challenge (Fig. 6D). The proportion of donor ResMØs, when expressed as a number relative to the presence in mice that did not receive zymosan, remained unchanged throughout (Fig. 6E). The majority of cells in the InfMØ gate were host derived (Fig. 6F). DNA content and pHH3 staining was recorded in both donor and host ResMØs during the inflammatory response. As illustrated (Fig. 6G), both ResMØ populations behaved in a similar fashion. With regard to pHH3 staining, we detected with a high degree of confidence a marked increase in the proportion of pHH3+ ResMØs (Fig. 6G). The background number of pHH3+ Res MØs was increased, presumably because of recovery from adoptive transfer injection-induced inflammation. Analysis of DNA content identified similar changes in DNA content, but the statistical power of this analysis was clearly much lower (Fig. 6G). DNA content analysis and pHH3 staining of InfMØs in this experiment showed minimal evidence of proliferation (data not shown).
We have conducted a study using definitive markers of in situ proliferation and ResMØs in polychromatic flow cytometry optimized for rare cell detection to monitor tissue MØ proliferation during development, adulthood and during an acute inflammatory episode. The accuracy of our selective identification of ResMØs during an inflammatory response was confirmed by the use of multiple selective surface antigens and by the subsequent adoptive transfer experiments. A small number of peritoneal ResMØs are bi- or multi-nucleate, whereas others can be visualized in the process of phagocytic clearance of apoptotic cells or other ill-defined matter (data not shown). Occurrence of mitotic events is very rare. Aware of these additional sources of cell-associated DNA and the problems this would cause with cell-cycle analysis, we sought a specific measure of MØ proliferation that could be used in polychromatic flow cytometry. Ki67 is restricted to the active phases of the cell cycle (G1, S, G2 and M) 15. Ki67 staining revealed a significant number of Ki67− MØs with >2N DNA content. Since Ki67 staining had relatively poor resolution, we also examined histone H3, which is selectively phosphorylated at a discrete stage of mitosis 16. pHH3+ ResMØs were rare in adult mice, but the specific timing of pHH3 in mitosis provided definitive identification of ResMØs actively in mitosis within the naïve tissue. The association of pHH3 staining with high levels of Ki67 antigen detection may reflect Ki67 relocation during mitosis 15. The occurrence of >2N DNA content in the absence of Ki67 expression casts doubt over the validity of isolated DNA content analysis for the study of mature ResMØ proliferation. However, the use of pHH3 staining provides a definitive marker of proliferative events occurring within the tissue.
Previous study of Langerhans cells 10 and to some extent peritoneal MØs 19 indicated that tissues may be populated by the local proliferation of ResMØs. Using pHH3 and DNA content to detect proliferation, we examined the proliferation of peritoneal MØs during ontogeny. The data indicated a marked increase in the proliferation of mature ResMØs in the young (2-wk old) mice, at a time when the numbers of ResMØs are greatly increasing as the animal develops. As the number of ResMØs stabilized in the adult, the proliferation decreased. Comparison of the proportions of cells with >2N DNA content or that are pHH3+in older animals again emphasized the problems of analyzing DNA content as a measure of cell cycle in ResMØs. The high level of mitosis in the young mice at a time of marked ResMØ expansion indicates that ResMØ numbers in the peritoneum are established by local proliferation.
During zymosan peritonitis, ResMØs are initially depleted and large numbers of InfMØs are recruited as part of the inflammatory process. The disappearance of ResMØs is primarily thought to represent firm adherence within the tissue and emigration to the draining lymph nodes 20, 21, many of these cells would most likely be permanently lost from the tissue. After the acute phase of inflammatory response, ResMØs are again recoverable. Aliphatic dye labelling experiments have indicated that those ResMØs present after acute inflammation are the same as those present before 22, but these and related data are over-interpreted, as the dyes could, for example, be transferred during the act of phagocytosis. We specifically examined young adult mice and studied DNA content alongside pHH3 staining in both ResMØs and InfMØs during acute zymosan-induced peritonitis. There was a clear burst in proliferation of the ResMØs, but not InfMØs, approximately 48 h after the induction of peritonitis. Importantly, this occurred in the presence of a substantial number of InfMØs, which are derived from monocytes. The proportion of pHH3+ ResMØs during resolution of inflammation was approximately a third of that observed in a fast growing immortalized cell line, indicating a substantial number of ResMØs are actively proliferating in the tissue. We used BrdU in the drinking water during the first 72 h after zymosan-induced peritonitis, to label those cells generated by proliferation. A significant number of ResMØs selectively incorporated BrdU during this period. In the previous experiments, in which we fed BrdU in the period of 72–168 h post-zymosan, we noted incorporation in just over 10% of ResMØs, with higher uptake in subsets of InfMØs that were present 168 h after zymosan-induced peritonitis 2. Combining the findings of these studies indicates that during an acute self-resolving inflammatory insult the majority of ResMØs are replaced by proliferation, with proliferation largely restricted to a specific time window. On the contrary, InfMØs incorporate BrdU later in the inflammatory response, which could be expected since pre-existing monocytes (BrdU−) are among the first cells recruited to the peritoneal cavity (see below also). BM and blood monocytes exhibited maximal labelling with 3H-thymidine 24 and 48 h after induction of peritonitis 6 consistent with this interpretation.
To determine if ResMØs from naïve mice were surviving inflammation and undergoing a proliferative burst to reconstitute the tissue after acute inflammation, we adoptively transferred cells from naïve donor mice to hosts with a different CD45 allotype. The donor ResMØs behaved exactly as the host ResMØs, experiencing the MØ disappearance reaction and then recovering again in numbers while experiencing an identical proliferative burst. At present, we are unable to exclude retention of adoptively transferred ResMØs in a specific location (e.g. the milky spots of the omentum), which are subsequently released after inflammation. Whatever the dynamics of possible additional pools of ResMØs, these, like the peripherally derived InfMØs, had negligible impact on the relative persistence of the adoptively transferred ResMØs (Fig. 6E), indicating little or no contribution to the ResMØs pool, and that local proliferation of ResMØs is the major mechanism of recovery from inflammation. Our data are very complementary to the similar conclusions that have been obtained from the study of a chronic parasite infection model 13.
Our data are consistent with a model in which ResMØ proliferation populates the tissue of the newborn. This is similar to what has been seen in the epidermis 10. Subsequently, when homeostasis is established in adult tissue, proliferation is limited to sustain the population. Restriction of ResMØ proliferation may continue during acute inflammation (perhaps by different mechanisms); however, we have demonstrated that once this has subsided without normal homeostatic controls, ResMØs are allowed to proliferate until homeostatic tissue levels of ResMØs are restored. In this context, peripheral cells may only contribute significantly to the ResMØ population after a critical failure of the ResMØs, such as substantive ablation of Langerhans cells by UV exposure 23. Proliferation of MØs has not been readily reported in humans, although the treatment of human monocytes with M-CSF results in regulation of genes involved in cell cycle 24 and a subset of monocytes can proliferate 25. Additionally, proliferation of MØs has been observed in renal disease 26, 27.
In summary, we have defined the mechanism for the maintenance and renewal of tissue MØs in a vascular tissue as one with controlled self-renewal under homoeostatic conditions, followed by recovery after inflammation with an intense period of proliferation. This proliferation occurs after a time when the tissue is loaded with inflammatory recruited MØs, which persist during the burst of proliferation of the ResMØs, and yet themselves show limited evidence of proliferation or contribution to the ResMØ population. This establishes that the model proposed for Langerhans cell and microglial renewal can be extended to vascular tissues and perhaps interpreted as a general paradigm for maintenance and renewal of tissue MØs, although this will still need to be evaluated on a case-by-case basis, considering inflammatory stimuli, dose and micro-environment.
Materials and methods
Mice and primary cells
All mice used in this study were 2- to 16-wk-old 129S6/SvEv or 129S6/SvEv.CD45.1 congenics (from our own colonies; the CD45.1 congenics being a kind gift from Professor F. Powrie, Oxford) or 6- to 7-wk-old C57BL/6 (Harlan or Charles River).
Peritoneal cells were recovered by lavage from both inflamed and naïve mice using 1.5–5 mL (depending on size of animal) of 5 mM ice-cold EDTA in PBS and a 21–23 gauge needle. Cells were stored on ice prior to analysis. Cytospin preparations were typically made using between 1 and 5×104 cells, spun at 500 rpm for 3 min. Cytospun cells were air dried, stained with Microscopy Hemacolour cell stain (Merck), visualized on a Leica DMLB microscope (Leica) and images were captured using a Leica DFC490 digital camera (Leica) and processed using QWin Software (Leica).
Conditionally immortalized MØ progenitors from 129S6/SvEv mice (MØ P-S6) 17 were cultured in RPMI 1640 medium containing 10% heat-inactivated fetal calf serum, 50 U/mL penicillin, 50 μg/mL streptomycin, 1 μM β-estradiol (Sigma) and 10 ng/mL GM-CSF (Peprotech). These cell lines are fast growing 28 and were used as a control for the efficiency of pHH3 detection.
Experimental inflammatory models
Zymosan-induced peritonitis was achieved by i.p. injection of 2×106 zymosan particles (∼10 μg) in 100 μL of PBS. In selected experiments, the mice were given drinking water containing 0.8 mg/mL BrdU (Sigma). At defined time points, animals were humanely sacrificed and the cells were recovered by peritoneal lavage. For adoptive transfer experiments, naïve donor 129S6/SvEv.CD45.1 congenic mice were humanely sacrificed and peritoneal cells were recovered by lavage with 1 mL of ice-cold PBS. To minimize manipulation of the donor cells, they were pooled from several donors and 0.5 mL was immediately i.p. injected into host 129S6/SvEv mice (CD45.2+). The MØ composition and cell count of the adoptively transferred cells were then determined retrospectively by flow cytometry using residual cells (see Flow cytometry). Typically, 1 day after transfer, 20–30% of F4/80hiCD11bhi ResMØs were donor-derived.
Flow cytometry was performed according to the conventional protocols on ice, in the dark and in the presence of 2 mM NaN3. The cells were blocked for 30 min with 0.5% w/v BSA, 5% v/v heat-inactivated rabbit serum, 4 μg/mL rat anti-mouse 2.4G2 (anti-FcγRII and -III) in PBS before the addition of primary antibodies. Primary antibodies were left for 1 h on ice before washing the cells three times with wash buffer (as block, but without serum or 2.4G2). For intracellular flow cytometry, cells were initially fixed in 1% formaldehyde (Sigma) for 20 min on ice before blocking, and 0.5% saponin (Sigma) was included in the block and wash buffers. The FITC 5-bromo-2′-deoxyuridine (BrdU) flow kit (BD) was used to detect BrdU incorporation. In the case of BrdU detection, the staining protocol was in accordance with the manufacturer's recommendations, except that surface antigens were first labelled as described above prior to fixation and permeabilization. When required, permeabilized cells were stained with 4′,6-diamidino-2-phenylindole (DAPI) (Invitrogen) or DRAQ5 29 (Biostatus) for >30 min before analysis. The samples were acquired using the Beckman-Coulter Cyan-ADP (3-laser) flow cytometer. The data were analyzed with the Summit (Beckman-Coulter) or FlowJo (Treestar) software. In all cases, data were displayed on logarithmic scales, except for the analysis of DNA content, which was on a linear scale. Doublet discrimination was always performed by concurrent gating on FSCarea versus FSClin, pulse width and, when used, DNAarea versus DNAlin as shown in Fig. 2. Measurements of the stage of cell cycle by DNA content were achieved using FlowJo software and the Watson pragmatic model. Measurements of proportion of cells that had incorporated BrdU were determined by Overton channel-by-channel subtraction of histograms using staining of mice not fed BrdU as a background staining control in Summit software. When cells were stained after fixation and permeabilization, eosinophils were excluded from the analysis based on their characteristic of FSC, SSC, F4/80+, CD11b+ profiles 3, and are not shown on plots. Neutrophils were ruled out by their F4/80−, Ly-6G+, small SSC, and FSC profiles after fixation and permeabilization. The markers, F4/80 and CD11b, were primarily used to select for the ResMØ profile, but Tim4, which labels the majority of ResMØs was used whenever possible to confirm this.
The following antibodies were used in this study: F4/80-PE, F4/80-allophycocyanin (APC), and F4/80-AlexaFluor (AF) 405 from Serotec (all clone CI:A3-1). CD11b-allophycocyanin-cyanine(Cy)7 (CD11b-allophycocyanin-Cy7) and CD11b-PerCP-Cy5.5 (both clone M1/70), Ly6G-PE-Cy7 (clone 1A8), Ki67-PE set (clone B56), and CD45.2-PerCP-Cy5.5 (clone 104) from BD PharMingen. pHH3-AF488 (Ser10) (clone D2C8) and rabbit IgG-AF488 (clone DA1E) from New England Biolabs. The Ly6B (clone 7/4) antibody was produced in house and labeled using the Phycolink PerCP-labeling kit (Europa Bioproducts). The CD45.1-allophycocyanin (clone A20) antibody was from eBioscience. The Tim4-PE (clone RMT4-54) antibody was purchased from Cambridge Bioscience.
Statistical analyses were conducted using the GraphPad Prism. The statistical tests used are indicated as appropriate within the text. p-Values are summarized as follows: *p<0.05, **p<0.01 and ***p<0.001.
P. R. T. is a Medical Research Council UK (MRC) Senior NonClinical Fellow (G0601617) and L. C. D. is an MRC and Cardiff University School of Medicine funded Ph.D. student and a Cardiff University 125 for 125 Scholar. All animal work was conducted in accordance with Institutional and UK Home Office guidelines. The authors thank the staff of our animal facility for the care of the animals used in this study.
Conflict of interest: The authors declare no financial or commercial conflict of interest.