Unveiling skin macrophage dynamics explains both tattoo persistence and strenuous removal

Recent studies, based on ontogeny and transcriptomic analysis, have succeeded in discriminating conventional dendritic cells (cDCs) from monocyte-derived cells. cDCs comprise two lineages known as cDC1 and cDC2 (Guilliams et al., 2016), have a short life span, and are continually replenished from committed cDC precursors that originate in the bone marrow (BM). Under steady-state conditions, BM-derived blood Ly-6C^high monocytes extravasate at a slow pace into nonlymphoid tissues, such as the intestinal lamina propria and the skin dermis, where they give rise to two developmental streams (Tamoutounour et al., 2012, 2013; Bain et al., 2013; Jakubzick et al., 2013). Progression through the first developmental stream is associated with the preservation of CC-chemokine receptor 2 (CCR2) expression, the acquisition of MHCII molecules, and a diminution of Ly-6C expression. It can be visualized in Ly-6C–MHCII dot plots as a “monocyte waterfall,” the final product of which share features with cDC2 and is called monocyte-derived DC (mo-DCs; Tamoutounour et al., 2013). Blood Ly-6C^hi monocytes extravasate at a higher pace into inflamed tissues (Tamoutounour et al., 2012, 2013; Bain et al., 2013) and give rise to cells that, because of their ability to produce TNFα and inducible NO synthase, are denoted as Tip-DCs (Serbina et al., 2003; De Trez et al., 2009; Tamoutounour et al., 2012). Under steady-state conditions, extravasated Ly-6C^high monocytes also proceed through a second developmental stream yielding tissue-resident macrophages (mo-Mac) that have lost expression of CCR2 and Ly-6C and express or not MHCII molecules (Jakubzick et al., 2013; Tamoutounour et al., 2013; Yona et al., 2013; Bain et al., 2014; McGovern et al., 2014). In the intestine and dermis, such mo-Mac replace in part or in totality a pool of macrophages that develops before the establishment of definitive hematopoiesis in a CCR2-independent manner (Tamoutounour et al., 2013; Bain et al., 2014; Sheng et al., 2015). Monocyte fate toward mo-Mac or mo-DCs is specified by microenvironmental cues (Goudot et al., 2017).

The possibility of discriminating cDCs from mo-DCs and macrophages emphasized their distinct functional specialization. Upon infection, antigen-laden cDCs migrate to draining LNs and trigger the conversion of antigen-specific naive T cells into effector T cells. In contrast, most mo-DCs remain in tissue parenchyma where they contribute to local immune responses and provide potent antimicrobial function during infection. Macrophages contribute to tissue homeostasis and wound repair because of their capacity to phagocyte cellular debris, invading organisms, and apoptotic cells. Owing to their strategic positioning at body barriers, macrophages capture a wide range of exogenous particulates. They also ingest self-derived particulates, as exemplified by the melanin-laden macrophages that have been sporadically observed in the normal human dermis and referred to as “melanophages” (Haniffa et al., 2009, 2012).

To study the functional specialization that exists among cells of the mononuclear phagocyte system, several strains of gene-targeted mice have been developed to ablate a given cell subset by placing the human diphtheria toxin receptor (hDTR) under the control of a gene promoter specific for the subset of interest. Those models are sometimes associated with drawbacks (van Blijswijk et al., 2015; Durai and Murphy, 2016). For instance, in CD11b-DTR mice, cDC2s, macrophages, and mo-DCs were depleted, and because of off-target DTR expression, those mice died after a few diphtheria toxin (DT) injections (Duffield et al., 2005). Although CCR2-DTR mice were not prone to DT-induced lethality (Hohl et al., 2009), some of the phenotypes observed by using those mice likely resulted from the depletion of CCR2⁺ cDC2s rather than of monocytes and monocyte-derived cells as generally assumed (Malissen et al., 2014).

In contrast to cDCs, both the mo-DCs and mo-Mac, including their precursors, and the macrophages of embryonic origin that are found in tissues such as the skin, the lung, and the gut express the high-affinity IgG receptor FcgRI (CD64; Bain et al., 2013; Plantinga et al., 2013; Tamoutounour et al., 2013). Here, we developed mice in which a hDTR gene was placed under the control of the Fcgr1 gene that codes for CD64. This model has a tighter ablation specificity compared with previous models in that it permitted specifically ablating macrophages, monocytes, and monocyte-derived cells in vivo. It allowed us to identify in the ear skin dermis of C57BL/6 (B6) mice a population of melanophages that constitutes a quantitatively major subset of dermal myeloid cells and has been missed in previous studies. By characterizing their phenotype, origin, and dynamics, we showed that they corresponded to dermal macrophages that have ingested melanin granules. Benefiting from the knowledge gained on melanophage dynamics, we further elucidated the identity, origin, and dynamics of the myeloid cells that are found in the mouse skin and capable of capturing and retaining tattoo pigment particles.

To assess the dynamics and functional role of CD64⁺ macrophages and monocyte-derived cells in vivo, we generated a line of gene-targeted mice containing in the 3′-UTR of the Fcgr1 gene a cassette composed of an internal ribosomal entry site, an enhanced green fluorescent protein (EGFP), a 2A cleavage sequence, and an hDTR. To demonstrate that the mo-Mac and monocyte-derived cells found in mice homozygous for the Fcgr1-IRES-EGFP-hDTR allele (also known as B6-Fcgr1^tm1Ciphe and called CD64^dtr mice here) developed normally and expressed normal levels of CD64, we analyzed their skin dermis because it contains a full complement of such cells (Fig. S1 A; Tamoutounour et al., 2013). Among dermal CD11b⁺ cells, cDC2s can be distinguished from non-cDC2s cells by their Ly-6C^–CD64^– phenotype. Dermal CD11b⁺CD64⁺ non-cDC2 cells can be further resolved into CCR2⁺ and CCR2^– cells. CCR2⁺ cells comprise Ly-6C^highMHCII^– dermal monocytes (P1), Ly-6C^high (P2), and Ly-6C^– (P3) MHCII⁺ mo-DCs that constitute a P1 → P2 → P3 developmental series. CCR2^– cells comprise MHCII^low (P4) and MHCII^high (P5) dermal macrophages. The ear skin dermis of DT-untreated CD64^dtr mice showed ratios and numbers of monocytes, mo-DCs, and macrophages comparable to those of WT mice (Fig. S1 B), and all those cells expressed normal levels of CD64 (Fig. S1 C). Similar results were obtained when CD64⁺ cells found in the lamina propria of the large intestine were analyzed (not depicted).

CD64^dtr mice were treated with DT twice and 24 h apart without lethality, and the myeloid cells present in the ear skin were analyzed at various time points after the last DT injection. DT treatment was not associated with overt inflammation as documented by the lack of skin infiltrating neutrophils (not depicted). CD64⁺CCR2⁺ cells were all ablated within 24 h (Fig. 1, A and B), demonstrating that the hDTR was functional. 48 h after the last DT injection, Ly-6C^highMHCII^– dermal monocytes started to return, and Ly-6C^high and Ly-6C^– MHCII⁺ mo-DCs were replenished in sequence over the next 5 d. A transient rebound in their numbers occurred around day 4, suggesting that their rate of generation can increase in response to acute, DT-induced loss. Both MHCII^low and MHCII^high CCR2^–CD64⁺ dermal macrophages showed protracted reconstitution kinetics compared with CD64⁺CCR2⁺ cells in that it took up to 10 d to fully reconstitute their pools (Fig. 1, C and D). Analysis of the gate corresponding to CCR2^–CD64⁺ cells at days 2–3 revealed a population of Ly-6C^highMHCII^{low to high} cells (denoted as Px cells in Fig. 1, C and D) that started losing Ly-6C expression on day 4 and thus progressively merged within gates corresponding to MHCII^low and MHCII^high dermal macrophages. The phenotype, kinetics of reappearance of Px cells, and the occurrence at days 4 and 5 of cells straddling the Px and P4/P5 gates suggest that they correspond to intermediaries between Ly-6C^high blood monocytes and MHCII^low and MHCII^high dermal macrophages. The percentage of Px cells among CCR2^–CD64⁺ cells reached pre-DT-treatment levels upon replenishment of the MHCII^low and MHCII^high dermal macrophage compartments (Fig. 1, C and D).

In the lamina propria of the large intestine of CD64^dtr mice, CD64⁺ cells were also ablated within 24 h after the last DT injection and reconstituted from 48 h onward akin to dermal CD64⁺ cells (Fig. S2). Therefore, by allowing to conditionally empty the niche occupied in the skin dermis and intestinal lamina propria by CD64⁺ myeloid cells, the CD64^dtr mouse model demonstrated that all the mo-DC and macrophage subsets that were present in adult skin and intestine before DT treatment can be readily reconstituted in normal ratios and numbers and with kinetics consistent with their expected precursor-product relationship.

Adult steady-state monopoiesis occurs in BM, where monocyte–macrophage DC progenitors (MDPs) give rise to common monocyte progenitors (cMoP) that generate two main blood monocyte subsets that are both CD115 (CSF1R)⁺ and can be distinguished on the basis of CD43 expression (Fig. 2 A). CD43^– cells are CX3CR1^intCCR2⁺CD64⁺Ly-6C^high and CD43⁺ cells are mostly CX3CR1^highCCR2^–CD64⁺Ly-6C^low (Sunderkötter et al., 2004). They are denoted here as Ly-6C^high and Ly-6C^low monocytes, respectively. Ly-6C^low monocytes remain within blood vessels (Carlin et al., 2013) whereas Ly-6C^high monocytes contribute to the replenishment of mo-DCs and mo-Mac in tissues, including the skin and the intestine (Jakubzick et al., 2013; Tamoutounour et al., 2013; Yona et al., 2013; Bain et al., 2014; McGovern et al., 2014). Intraperitoneal injection of DT twice and 24 h apart resulted within 24 h in the complete ablation of the MDPs and cMoPs and Ly-6C^high, Ly-6C^int, and Ly-6C^low monocytes found in the BM (Fig. S3) and of the Ly-6C^high and Ly-6C^low blood monocytes (Fig. 2, B–D). Analysis of the blood and BM of CD64^dtr mice at various time points after the last DT injection showed that blood Ly-6C^high and Ly-6C^low monocytes were rapidly replenished and that a transient rebound in their numbers occurred (Fig. 2, B–D; and Fig. S3). Consistent with the observation that CD115 is dramatically down-regulated in response to CSF-1 (Yu et al., 2012), the earliest BM-derived monocytes reappearing in the blood expressed lower levels of CD115 than CD115⁺ steady-state monocytes and are denoted here as CD115^int cells (Fig. 2, A, B, and E). Because of their CD115^int phenotype, those Ly-6C^high monocytes were not recorded in the CD115⁺CD43^– gate used in Fig. 2 C. Their numbers peaked at day 3 after DT treatment (Fig. 2, B and E), and when blood monocytes were fully replenished, their relative representation dramatically decreased. Congruent with the view that circulating CD43^–Ly-6C^high monocytes give rise to CD43⁺Ly-6C^low monocytes (Varol et al., 2007; Thomas et al., 2016; Mildner et al., 2017), the first CD43⁺ monocytes reappeared at days 3–4 after the last DT treatment and gradually lost Ly-6C expression (Fig. 2 D). An approximately fourfold increase in blood neutrophils occurred in CD64^dtr mice and peaked 3 d after DT treatment (Fig. 2 B). It occurred, however, with a far lower magnitude and shorter duration than that observed upon DT administration in CD11c-DTR mice (Autenrieth et al., 2012; Tittel et al., 2012).

We analyzed next whether DT treatment of CD64^dtr mice affects the cDCs found in the skin. Langerhans cells (LCs; CD24^hiCD11b⁺), cDC1s (CD11b^–CD24⁺), cDC2s (CD24^lowCD11b⁺), and double-negative (DN) cDCs (CD11b^–CD24^–) were identified as described in Fig. S4 A. Although the absolute numbers of LCs, cDC1s, and DN cells were not affected by DT treatment, a 1.3-fold reduction in cDC2 absolute number was observed 1 d after DT treatment (Fig. S4 B). Therefore, our regimen of DT administration had a rather minimal impact on the cDC subsets found in the skin of CD64^dtr mice. CD64^dtr mice can thus be used to monitor the dynamics and function of macrophages and monocyte-derived cells without major collateral effects or lethality.

Former analyses of CD45⁺Lin^– myeloid cells present in the ear skin dermis focused on SSC-A^lowFSC-A^{low to high} (SSC-A^low in short) cells, leaving aside an uncharacterized population of SSC-A^highFSC-A^{low to int} (SSC-A^high in short) cells (Tamoutounour et al., 2013). After treatment of CD64^dtr mice with DT twice and 24 h apart, all those SSC-A^high cells were unexpectedly ablated (Fig. 3 A). Further characterization of such DT-sensitive, SSC-A^high cells showed that they had a CD24^lowCD11b⁺CD64⁺CCR2^–MHCII^{+ or –} phenotype identical to that of SSC-A^low dermal macrophages and that CD64 expression accounted for their DT sensitivity (Fig. 3 B).

In the almost hairless skin of mouse ears, melanocytes are primarily found in the dermis (Fig. 3 C; Aoki et al., 2009) and synthetize melanin pigment within specialized lysosome-related structures known as melanosomes. Considering that the SSC-A parameter used in cytometry is proportional to cell granularity and that CD64^dtr mice were of B6 origin and have a black coat, we surmised that the high granularity of the SSC-A^high myeloid cells found in the ear dermis was due to their capture of melanin granules from neighboring melanocytes. Hematoxylin and eosin staining of cytospin preparations showed that the SSC-A^lowCCR2^–CD64⁺ cells and the SSC-A^high cells sorted from the ear dermis have typical features of tissue macrophages; they were large with prominent cytoplasmic vacuoles (Fig. 3 D). The vacuoles of SSC-A^high cells differed, however, from those of SSC-A^low CCR2^–CD64⁺ cells in that they were loaded with dark material (Fig. 3 D). Upon analysis by electron microscopy, such dark vacuolar material had an extremely electron-dense appearance and an ellipsoidal shape (∼0.2 µm × 0.7 µm), two hallmarks of melanosomes (Fig. 3 E). All the ingested melanosomes were in a mature stage, showing a uniformly electron-dense granule without discernable internal structure (Fig. 3 E).

In contrast to their precursors, dermal macrophages do not express CX3CR1 (Tamoutounour et al., 2012). Consistent with our hypothesis that melanophages correspond to macrophages that ingested melanosomes, they were CX3CR1^– (not depicted) and identified as melanin⁺tdTomato⁺CD64⁺ cells in the ear section of CX3CR1^Cre/+ Rosa^lsl-tdt/+ mice that fate map cells that have expressed CX3CR1 during their development (Fig. 3 F). In the densely haired mouse trunk, melanocytes are absent from the dermis and confined to the hair follicles where melanosomes are transferred from melanocytes to matrix keratinocytes to give rise to melanized hairs (Fig. 3 C). Such transfer is strictly coupled to the anagen phase of hair growth, and melanin is not formed during the telogen phase (Fig. 3 C; Slominski et al., 2005). Consistent with these anatomical features, the CD11b⁺CD64⁺CCR2^– macrophages isolated from the trunk skin during both the anagen and telogen phases lacked melanophages (Fig. 3 G). Likewise, the ear skin of B6-albino and BALB/c mice, which is deprived of melanic pigment, also lacked SSC-A^high melanophages (Fig. 4, A and B). Therefore, the presence of dermal melanophages correlates with the presence of dermal melanocytes. Although they represent 55 ± 10% of the monocyte-derived cells present in the B6 ear dermis (Fig. 4, A and B), it is likely that melanophages have been missed in previous studies because of their unusual SSC-A^high phenotype.

To assess the dynamics of dermal melanophages replenishment after DT-induced depletion, CD64^dtr mice were treated with DT twice and 24 h apart. Melanophages were all ablated within 24 h, and their numbers fully reconstituted ∼20 d after the last DT injection (Fig. 5, A and B). The granularity of SSC-A^high cells, a proxy for melanin content, increased with time. The upper section (>190 K) of the SSC-A^high gate was only filled between days 20 and 90 after DT treatment, and a melanin content distribution identical to pre-DT-treatment levels was regained 90 d after DT treatment (Fig. 5, A and B). Although the dermis of CCR2-deficient mice is only populated by macrophages of prenatal origin that self-maintain without any input from adult monopoiesis (Tamoutounour et al., 2013), it showed a melanophage to macrophage ratio and melanin content distribution identical to that of WT dermis (Fig. 6 A). Therefore, the macrophages found in the ear dermis have the same propensity to capture melanin regardless of their prenatal (Fig. 6 A) or adult (Fig. 5) origin.

To demonstrate without relying on DT ablation that replenishment of the pool of adult dermal macrophages and melanophages relied on continuous input from BM-derived circulating precursors, parabiosis was established between adult mice expressing CD45.1 and CD45.2 (Fig. 6 B). Blood monocytes were allowed to exchange between parabionts for 8 wk before assessing the presence of donor-derived macrophages and melanophages in the ear skin of each parabiont. Dermal MHCII⁺ macrophages contained ∼15% of cells originating from the parabiotic partner (Fig. 6 B), a distribution comparable to that of Ly-6C^high blood monocytes and demonstrating that they continuously develop from BM monocytes (Tamoutounour et al., 2013; Sheng et al., 2015). Consistent with the view that a fraction of dermal MHCII^– macrophages is of embryonic origin and self-renews in situ (Sheng et al., 2015), MHCII^– macrophages showed a twofold reduced rate of exchange with blood-borne precursors compared with dermal MHCII⁺ macrophages. In parabiosis experiments involving WT and CCR2-deficient mice, the percentages of donor-derived cells found in the CCR2-deficient parabionts reached 80% and 40% of MHCII⁺ and MHCII^– macrophages, respectively, compared with 80% for Ly-6C^high blood monocytes (Fig. 6 B). As expected for cells linked by a precursor-product relationship, in both types of parabiosis MHCII⁺ and MHCII^– melanophages achieved a degree of chimerism identical to that of MHCII⁺ and MHCII^– macrophages (Fig. 6 B). Altogether, these data indicate that dermal macrophages and melanophages are related through a precursor-product relationship and that dying dermal macrophages and melanophages are continuously replenished with new ones via CCR2-dependent recruitment of circulating Ly-6C^high blood precursors.

To assess whether melanophages show a unique transcriptomic signature compared with dermal SSC-A^low macrophages that have not ingested melanin, MHCII⁺ and MHCII^– melanophages and macrophages from the ear dermis were sorted in triplicates and subjected to gene-expression profiling. Principal component analysis (PCA) on all the genes found to be differentially expressed between at least two of the four subsets showed that the MHCII⁺ and MHCII^– status accounted for most of the overall variability observed in the dataset (47%) and that differences between melanophages and macrophages was only the second component of variability (17%; Fig. 7 A). Insights into the biological pathways underlying these transcriptomic differences were obtained through ingenuity pathway analysis (IPA) of eight lists of genes corresponding to those differentially expressed (1) in one out of the four subsets, (2) between MHCII⁺ versus MHCII^– cells, and (3) between melanophages and macrophages (Fig. 7 B). When compared with their MHCII^– counterparts, MHCII⁺ macrophages and melanophages were significantly enriched in annotations related to the adaptive immune response such as antigen presentation pathway, dendritic cell maturation, and cytokine signaling (IL-4, IL-6, IL-10, and IL-17). MHCII expression was also associated with a pro-inflammatory signature with higher expression of Ccr7, Il1b, Il1r2, and Il6 and to immune cell chemotaxis with higher expression of Ccl17, Ccl22, Ccl5, and Cxcl13 (Fig. 7 C). When compared with SSC-A^low macrophages, melanophages were significantly enriched in annotations related to aryl hydrocarbon receptor signaling and glutathione-mediated detoxification (Fig. 7 B) because of their expression of two genes of the glutathione S transferase family (Gstm1 and 3; Table S1). High and selective expression by melanophages of genes coding for a lysosomal ATPase and a cystein protease (Atp6v0d2 and Ctsk, respectively) led to their enrichment for the annotation phagosome maturation (Fig. 7 B and Table S1), a feature consistent with the fact that they are laden with melanosomes.

Tumors depend on the supportive functions of macrophages. Clusters of melanophages have been observed in the periphery of the nevi found in the dermis of mice expressing oncogenic mutated form (BrafV600E) of the serine-threonine-specific protein kinase BRAF (Dhomen et al., 2009). To determine whether such melanoma-associated melanophages correspond to CD64⁺SSC-A^high cells, B6-albino mice, which lack melanophages, were injected with melanocytic (B16) or amelanoctyic (Braf^V600E) melanoma cells. Analysis of CD45⁺Lin^–Siglec-F^– cells isolated from tumor masses 14 to 18 d after inoculation showed that SSC-A^high cells were present only in the melanocytic melanoma (Fig. 8 A). Their morphology was similar to that of the melanophages identified in the B6 dermis (Fig. 8 B) and they had a CD11b⁺CD64⁺F4/80⁺ phenotype (Fig. 8 C).

The melanophages that are identifiable in melanocytic melanoma constitute a unique thread permitting revisiting the phenotype of tumoral macrophages. For instance, the presence of CD64 and absence of Ly-6C permit readily distinguishing macrophages in steady-state tissues (Tamoutounour et al., 2013), and it has been assumed that such a feature holds true within the tumor microenvironment (Salmon et al., 2016). Our analysis of the melanophages found in melanocytic tumors showed, however, that most of them expressed Ly-6C (Fig. 8 C). Therefore, analysis of the melanophages found in melanoma constitute a unique decision support tool permitting to determine whether markers defined as macrophage-specific under steady-state conditions keep or lose their discriminatory potential within the tumor microenvironment.

Using the CD64^dtr mouse model and benefiting of our knowledge on melanophages, we determined next the identity, origin, and dynamics of the skin myeloid cells that are capable of capturing and retaining tattoo pigment particles. The frequency of melanophages in the tail skin of B6 mice is fourfold lower than that of melanophages in ear dermis and reflects its scarcity in melanocytes. Moreover, because of their reduced melanin content tail skin melanophages barely merged in the SSC-A^high gate, the lower limit of which that was defined by using B6 albino ear (Fig. 4, A and B). Accordingly, we used tail skin of B6 mice to monitor the appearance in the almost-empty SSC-A^high gate of those cells that have captured the bright green pigment particles that are contained in the tattooing ink paste used in the present study. Analysis of the percentages of SSC-A^low and SSC-A^high cells found among the CD45⁺Lin^– cells isolated from the skin of tails that were tattooed 3 wk before analysis showed that they contained a unique population of SSC-A^high cells compared with untreated mice (Fig. 9 A). Further characterization of those SSC-A^high cells showed that they had a CD24^lowCD11b⁺CD64⁺CCR2^{– to low} phenotype (Fig. 9 A), which was identical to that observed for SSC-A^high melanophages and SSC-A^low macrophages (Fig. 3 B). Cytospin analysis of SSC-A^high cells sorted from tattooed tail skin showed that they were large cells containing prominent cytoplasmic vacuoles loaded with green material, whereas SSC-A^low dermal macrophages and cDCs contained no green tattoo pigment (Fig. 9 B). None of the CD45^– skin cells, which mostly comprise keratinocytes, contained green material (not depicted). Therefore, 3 wk after tail tattooing, the green pigment particles present at the site of tattooing were primarily found within dermal macrophages.

We used CD64^dtr mice to ablate dermal CD64⁺ macrophages laden with green pigment particles (“green” macrophages in short) and to determine whether their death results in tattoo vanishing. Accordingly, CD64^dtr mice that had the skin of their tail tattooed with green tattoo ink paste 3 wk before were treated with DT twice and 24 h apart. Analysis of tattooed skin of those mice for the presence of SSC-A^high green macrophages showed that they were as expected all ablated within 2 d after DT treatment and that, akin to melanophages, their pool was slowly reconstituted after DT injection (Fig. 9 C). Their granularity, reflecting green-particle content per cell, increased with time and 90 d after DT injection reached levels identical to those present just before DT treatment (Fig. 9, A and C). The pool of SSC-A^high cells that was replenished 90 d after DT treatment was exclusively composed of green macrophages with a morphology and CD24^lowCD11b⁺CD64⁺CCR2^{– to low} phenotype resembling those of green macrophages observed pre-DT treatment (Fig. 9 D and not depicted).

Considering that no green tattoo ink paste was reapplied after DT treatment, it is likely that most of the green pigment particles released after DT-induced green macrophage death remained in an extracellular form at the site of tattooing where they were progressively recaptured by incoming dermal macrophages originating from circulating Ly-6C^high blood monocytes (see model A in Fig. 10). Consistent with that model, cytospin preparation of the total cell suspension from tail skin showed that before DT treatment green pigment particles were all associated with cells that had typical features of tissue macrophages, whereas 2 d after the last DT injection no green macrophages were left (Fig. 9 C) and the green pigment particles that were spun down with the cells were all extracellular (Fig. 9 E). Importantly, macroscopic examination of tattooed tails showed that even at time points when green macrophages were absent and green pigment particles were extracellular (for instance, day 2 after DT), the tattoo kept the appearance it had just before DT treatment (Fig. 9 F). Therefore, despite the fact that the green tattoo pigments went, over 90 d, through a full cycle of capture–release–recapture, the tattoo kept its macroscopic appearance.

To further support our pigment capture–release–recapture model without relying on DT ablation, CD45.1⁺ B6 mice were tattooed on the tail. The tattooed tail skin was explanted and grafted on the skin of the back of CD45.2⁺ B6 albino mice. Analysis of the myeloid cells present in the graft 6 wk after transplantation showed that most of the SSC-A^high cells found among CD11b⁺CD64⁺CCR2^– cells were of host origin, suggesting that green tattoo pigments were released from the CD45.1⁺CD11b⁺CD64⁺CCR2^–SSC-A^high cells originally present in the donor graft and captured by incoming CD45.2⁺CD11b⁺CD64⁺CCR2^– macrophages of host origin (Fig. 9 G). As discussed below, those findings provide a cellular basis for long-term tattoo persistence and have implications for improving strategies aimed at removing tattoos.

We used a novel model of DT-induced depletion of macrophages to generate niche availability without triggering inflammation and showed that the pool of adult dermal macrophages can be fully reconstituted from circulating Ly-6C^high monocytes. By using a model of DT-mediated depletion of liver Kupffer cells, circulating Ly-6C^high monocytes were also found capable of colonizing the emptied liver niches and of adopting a Kupffer cell identity (Scott et al., 2016). Likewise, transferred BM-derived monocytes were able to colonize the empty alveolar macrophage niche of neonatal CSF2RB-deficient mice and generate alveolar macrophages (van de Laar et al., 2016). Therefore, our data showed that the capacity for Ly-6C^high blood monocytes to replenish macrophages after DT-induced ablation also extends to the skin dermis. Moreover, we characterized the melanophages that are present in large numbers in the normal ear dermis of mice with a black coat, demonstrating that they correspond to dermal macrophages that have ingested melanosomes originating from neighboring melanocytes.

Limited information has been provided on the identity of the cells that are found in tattooed skin biopsies and capable of retaining tattoo pigment particles. By capitalizing on the knowledge gained on melanophages and on the functionalities offered by CD64^dtr mice, we conducted the first systematic analysis of the identity, origin, and dynamics of the skin myeloid cells that are capable of capturing and retaining tattoo pigment particles. The tattooing process consists of inserting water-insoluble pigment of the desired color into the dermis layer of the skin. Tattoo pigments are primarily metal salts in the submicrometer range. Immediately after tattooing, a fraction of pigments is shed with the epidermis or transported away from the dermis via lymphatic vessels. We demonstrated that the pigment particles that remain at the site of injection and cause the long-term tattoo color were exclusively found within dermal macrophages. Zaba et al. (2007) previously identified a population of cells located in the human dermis that was positive for factor XIIIA and the macrophage scavenger receptor CD163 that lacked CD11c and BDCA-1 and weakly activated allogenic T cells. Those cells have structural and ultrastructural features of macrophages and in biopsies of tattoos they were selectively laden with granular pigments. The fate of tattoo pigment injected into dermal tissues has been studied in the past, and fibroblasts were considered the primary long-term reservoir of the pigment granules (Ferguson et al., 1997; Elsaie et al., 2009). However, the morphological study of Zaba et al. (2007) and our present data combining multiparameter cytometry and the use of CD64^dtr “depleter” mice demonstrate that dermal macrophages constitute the primary store of dermal tattoo pigment in both mice and humans.

Although the dermis lacked macrophages for approximately 1 wk after DT treatment and it took several more weeks for the incoming macrophages to scavenge the released cell-free tattoo pigments, no macroscopic modification of the tattoo occurred. Before our study, the longevity of the undefined tattoo-pigment-laden cells found in the dermis was thought to be in the same order of that of a human adult, thereby accounting for long-term tattoo persistence (Fig. 10 B). Based on the present results, an alternative model for long-term tattoo persistence can be proposed. It takes into consideration the facts that the macrophages found in the dermis of adult mice are continuously replenished from circulating monocytes to compensate for the loss of dying macrophages and that most of the tattoo particles that are released from macrophages after DT treatment of CD64^dtr mice remain in situ and are recaptured by incoming macrophages. According to the pigment capture–release–recapture model (Fig. 10 A), in physiological conditions, when tattoo-pigment-laden macrophages die during the course of adult life, neighboring macrophages recapture the released pigments and ensure in a dynamic manner the macroscopic stability and long-term persistence of tattoos. Note that, in contrast to dermal cDCs, dermal macrophages do not have the capacity to migrate to draining LNs under normal and inflammatory conditions (Tamoutounour et al., 2013), an essential property for insuring tattoo persistence. Whether the pigment capture–release–recapture model of tattoo persistence applies to humans remains to be determined. In the case of the melanophages found in the human dermis, it has been suggested that they persist for years through longevity rather than continual renewal (Bigley et al., 2011).

Professional tattoos last for years or decades, and when no longer desired, they can be removed through the use of quality-switched lasers. Laser pulses lead to lysis of the tattoo-pigment-laden cells, fragmentation of the tattoo pigments, and their lymphatic transportation. Several cycles of laser treatment are required to achieve tattoo removal, and some tattoos remain immune to full removal. These difficulties are generally accounted for by the fact that a fraction of the fragmented pigment particles remains at the tattooing site and are recaptured by neighboring cells, a possibility formally demonstrated here by using the CD64^dtr model. Considering the long duration needed for the reconstitution of the pool of dermal macrophages and the recapture of the whole released pigments, tattoo removal can be likely improved by combining laser surgery with the transient ablation of the macrophages present in the tattoo area. As a result, the fragmented pigment particles generated by using laser pulses will not have the possibility to be immediately recaptured, a condition increasing the probability of having them drained away via the lymphatic vessels.

In conclusion, we have validated a new model of DT-mediated depletion of tissue-resident macrophages, monocytes, and monocyte-derived cells in which DCs are spared. By permitting the emptying of the niche occupied in the skin dermis and intestinal lamina propria by CD64⁺ myeloid cells, the CD64^dtr mouse model showed that all the mo-DC and macrophage subsets that were present in adult skin and intestine before DT treatment can be readily reconstituted from circulating Ly-6C^high monocytes. Furthermore, we showed that the myeloid cells of the ear skin dermis are dominated by melanophages, and we determined their origin and dynamics. Benefiting by our knowledge on melanophages, we determined the identity, origin, and dynamics of the skin myeloid cells that are capable of capturing tattoo-pigment particles and showed that they correspond exclusively to dermal macrophages. Based on those results, we proposed a dynamic model that accounts for long-term tattoo persistence and suggest that tattoo removal can be improved by combining laser surgery with the transient and specific ablation of the macrophages present in the tattooed area. Finally, we showed that melanocytic melanoma host readily identifiable melanophages that permit determining whether markers defined as macrophage-specific under steady-state conditions keep or lose their discriminatory potential within the tumor microenvironment.

Ccr2^−/− mice have been described (Jung et al., 2000). Cx3cr1^cre/+ (MW126Gsat) mice were generated by N. Heintz (The Rockefeller University, GENSAT, New York) and purchased from MMRRC. Rosa^lsl-tdt/+ (Ai14[RCL-tdT]-D) have been described (Madisen et al., 2010) and purchased from The Jackson Laboratory. Generation of Fcgr1-IRES-EGFP-hDTR gene-targeted mice is described below. C57BL/6 (B6) and BALB/c mice were purchased from Janvier and B6 albino (B6[Cg]-Tyr^c-2J/J) and CD45.1 C57BL/6 mice from Charles River. All mice, except BALB/c, were maintained on a B6 background. Mice were housed under specific pathogen–free conditions, and in vivo procedures were performed following protocols approved by the Ethics Committee of Marseille in accordance with institutional, national, and European directives for animal care (approvals APAFIS 779–2015 0605 10534083 and APAFIS 781–2015 0605 14006182).

A 6.2-kb genomic fragment encompassing exons 4 to 6 of the Fcgr1 gene (ENSMUSG00000015947) was isolated from a BAC clone of B6 origin (clone RP23-427P23; http://www.lifesciences.sourcebioscience.com). By using ET recombination, an IRES-EGFP-2A-hDTR-loxP-Cre-neoR-loxP cassette was introduced in the 3′ untranslated region of the Fcgr1 gene, 33 bp downstream of the stop codon. The targeting construct was abutted to a cassette coding for the DT fragment A expression cassette and linearized with Pme1. JM8.F6 C57BL/6N embryonic stem (ES) cells (Pettitt et al., 2009) were electroporated with the targeting vector. After selection in G418, ES cell clones were screened for proper homologous recombination by Southern blot. When tested on Sca1-digested genomic DNA, the 3′ single-copy probe used to identify proper recombination events hybridized to a 7.0-kb WT fragment and to a 10.3-kb recombinant fragment. When tested on HindIII-digested genomic DNA, the 5′ single-copy probe used to identify proper recombination events hybridized to a 10.7-kb WT fragment and to a 6.3-kb recombinant fragment. A neomycin-specific probe was used to ensure that adventitious nonhomologous recombination events had not occurred in the selected ES clones. Properly recombined ES cells were injected into FVB blastocysts. Germline transmission led to the self-excision of the loxP-Cre-NeoR-loxP cassette in male germinal cells. Fcgr1-IRES-EGFP-hDTR mice were identified by PCR of tail DNA. The pair of primers, sense 5′-CCCTTCCTCCCAGTGACAGTACTG-3′ and antisense 5′-TGAACCCATCCACCCTGTGAG-3′, amplifies a 402-bp band in case of the WT Fcgr1 allele, whereas the pairs of primers, sense 5′-CCCTTCCTCCCAGTGACAGTACTG-3′ and antisense 5′-CCTGCAAAGGGTCGCTACAGA-3′, amplified a 349-bp band in the case of the Fcgr1-IRES-EGFP-hDTR allele. Because of its low level of expression, the EGFP reporter was not used in the present study.

To isolate myeloid cells from ear skin, ears were processed as described (Bouladoux et al., 2017). In brief, ears were split into ventral and dorsal parts and incubated for 105 min at 37°C in RPMI containing 0.25 mg/ml Liberase TL (Roche Diagnostic Corp.) and 0.5 mg/ml DNase I (Sigma Aldrich). Digested tissue was homogenized by using Medicons and Medimachine (Becton Dickinson) to obtain homogenous cell suspensions. To isolate myeloid cells from tail skin, the skin was separated from the muscle and the vertebra and cut into small pieces and then digested in RPMI with Liberase TL and DNase as described for ear skin. The digested tissue was homogenized by using C Tubes and GentleMACS Dissociator (Miltenyi Biotec). To isolate myeloid cells from the lamina propria of the large intestine, the large intestine was opened longitudinally, cut into pieces, and washed in HBSS. The epithelial layer was eliminated by two incubations of 20 and 30 min in HBSS containing 2 mM EDTA. Intestinal pieces were washed twice in HBSS and then digested in RPMI medium containing 10% FCS, 1 mg/ml Collagenase 8 (Sigma Aldrich), and 30 ng/ml DNase I. Tissue was homogenized in C Tubes and GentleMACS Dissociator. Light-density cells were then purified by centrifugation on a 70–40% Percoll gradient (GE Healthcare). To isolate myeloid cells from tumor mass, tumors were cut into small pieces and then digested in RPMI medium containing 5% FCS, 0.8 mg/ml Collagenase 4 (Worthington), and 0.2 mg/ml DNase I for 40 min at 37°C. The digested tissue was homogenized by using C Tubes and GentleMACS Dissociator (Miltenyi Biotec).

Blood was collected in EDTA-containing tubes and then transferred to Trucount Absolute Counting Tubes (Becton Dickinson). Red blood cells were lysed with BD FACS Lysing Solution (Becton Dickinson). The absolute number of cells present in a given volume was determined by using the bead-to-cell ratio.

APC-Vio770–conjugated anti-NK1.1 (PK136), anti-CD3 (17A2), anti–Ly-6G (1A8), and anti-CD19 (6D5) were from Miltenyi Biotec. Biotin-conjugated anti–mouse-MERTK (BAF591) and PE-conjugated anti-CCR2 (475301) were from R&D. BV711-conjugated anti-CD11b (M1/70); BV605-conjugated anti-CD11c (N418); FITC-conjugated anti-CD43 (S7); Alexa Fluor 700–conjugated anti-CD45.1 (A20); Alexa Fluor 700–conjugated rat IgG2b κ isotype control (R35-38); FITC-conjugated anti-CD45.2 (104); APC-conjugated anti-CD64 (X54-5/7.1); FITC-, PE-CF594-, and BV421-conjugated anti–Ly-6C (AL-21); BV650-conjugated anti-NK1.1 (PK136); BV421-conjugated anti–Siglec-F (E50-2440); biotin-conjugated anti–Ly-6C (AL-21); and biotin-conjugated anti-CD103 (M290) were from BD PharMingen. PE-Cy5–conjugated anti-CD24 (M1/69), BV650-conjugated anti-CD45.1 (A20), PE-conjugated anti–Ly-6G (1A8), Pe-Dazzle–conjugated anti-CD45.2 (104), APC-conjugated anti-F4/80 (BM8), APC-conjugated rat IgG2a κ isotype control (RTK2758), APC-conjugated anti-CD11c (N418), APC-conjugated Armenian hamster IgG isotype control (HTK888), APC-conjugated anti-CD5 (53–7.3), APC-conjugated anti-CD19 (6D5), APC-Cy7–conjugated anti-CD11b (M1/70), FITC-conjugated anti-CD11b (M1/70), FITC-conjugated rat IgG2b κ isotype control (MRG25-8T), BV711-conjugated anti-CD64 (X54-5/7,1), BV711-conjugated mouse IgG1 κ isotype control (MOPC-21), and BV510- and Alexa-700–conjugated anti–MHC Class II (IA/I-E; M5/114.15.2) were from Biolegend. PE-Cy5.5–conjugated anti-CD45 (30-F11) and Pe-Cy7–conjugated anti-CD115 (AFS98) were from eBioscience. Biotin-conjugated antibodies were detected by using streptavidin conjugated with PE-CF594 or Pe-Cy7 (BD PharMingen).

Cells were stained and acquired on a FACS LSRII system and analyzed with FlowJo software (Tree Star). Neutrophils and B, T, and NK cells were systematically gated out by using a Lin⁺ gate (Fig. S1 and Fig. 2).

The different cell types were sorted on a FACS Aria II according to the marker combination specified on Fig. S1 A and Fig. 3. For microarrays (Affymetrix 1.0ST), cells were sorted directly in lysis buffer provided from the Qiagen RNAeasy PLUS extraction kit.

Sorted cells were spun onto Shandon Cytoslides with a Thermo Shandon Cytospin 4 cytofuge (4 min at 400 g), fixed with methanol and stained with hematoxylin and eosin.

Sorted cells were incubated at 37°C on poly-Lysine–coated cover slides. After 30 min, attached cells were fixed by using 2% glutaraldehyde in 0.1 M cacoldylate buffer, pH 7.4 for 1 h, postfixed for 1 h with 2% buffered osmium tetroxide, dehydrated in a graded series of ethanol solution, and then embedded in epoxy resin. Images were acquired with a digital camera Quemesa (SIS) mounted on a Tecnai Spirit transmission electron microscope (FEI Company) operated at 80 kV.

Trunk and ear skin pieces were fixed in Antigen Fix (Microm Microtech) for 2 h, washed in 0.1 M phosphate buffer, and dehydrated overnight in 30% sucrose in 0.1M phosphate buffer. Tissues were snap frozen, and 25-µm sections were stained with DAPI. To obtain an en face section of ear dermis, the dorsal part was split from the ventral one 2 h after intradermal ear injection of CD64-APC (2 µg), fixed in Antigen Fix for 2 h, washed in phosphate buffer, stained with DAPI, and clarified with Rapiclear 1.47 (SunJin Lab). Immunofluorescence confocal microscopy was performed by using a Zeiss LSM 880 confocal microscope. Separate images were collected for each fluorochrome and merged to obtain a multicolor image. Final image processing was performed with Imaris software (Bitplane) and Adobe Photoshop.

Transcriptomic and bioinformatics analyses were performed as previously described (Tamoutounour et al., 2013). To perform a PCA focused on all variability across the four biological samples examined (Fig. 7 A), we selected all ProbeSets showing a fold change ≥1.2 between at least one pair of conditions, taking into account only those ProbeSets nonambiguously annotated to one gene only and further keeping only the highest ProbeSet for each gene. This led to selection of 1,054 unique genes. To select the genes showing a significant differential expression in one type of condition compared with the other three, between MHCII⁺ versus MHCII^– cells, or between melanophages and macrophages (Fig. 7 B), we used the GeneSign module of the BubbleGUM software using the metrics Minimal pairwise (Mean[test]/Mean[ref]) with a linear fold change threshold of 1.5 and a P value threshold of 0.05. The microarray data have been deposited in the GEO database under accession number GSE103339.

Parabiotic mice were generated by using 8–9-wk-old male mice as described previously (Tamoutounour et al., 2013). For WT-WT parabiosis, CD45.1 B6 mice were sutured together with CD45.2 B6 mice, whereas for WT-Ccr2^−/− parabiosis, WT CD45.1 B6 mice were sutured with Ccr2^−/− CD45.2 B6 mice and subsequently treated with Bactrim for 8 wk before analysis.

CD64^dtr mice were injected intraperitoneally twice and 24 h apart with 1 µg DT (Calbiochem, EMD Millipore) in PBS. CD64^dtr mice injected with PBS showed no ablation of monocyte-derived cells and macrophages.

B6-albino mice were injected subcutaneously with 10⁵ amelanocytic (Braf^V600E) or melanocytic (B16) melanoma cells (Brown et al., 2001; Zelenay et al., 2015).

Tattooing was performed on the tail skin of mice under general anesthesia. Tail skin was tattooed by using the Animal Tattoo Ink Paste (Ketchum Mfg. Co, Inc.) and a 25-gauge needle. The ink-coated needle permits ink delivery in a topical manner. Before analysis, mice were kept for at least 3 wk until complete healing of the treated skin.

CD45.1 B6 mice were tattooed as described above. After 5 wk, the tattooed tail skin was explanted and grafted on the back of CD45.2 albino mice under general anesthesia. The back skin of the host was analyzed 6 wk after the graft.

The unpaired Student t test was used for statistical analyses with GraphPad Prism software.

Fig. S1 shows the gating strategy used to identify CD11b⁺ cell subsets within the skin of WT and CD64^dtr mice. Fig. S2 shows the ablation of CD64⁺ cells present in the lamina propria of the large intestine upon DT treatment of CD64^dtr mice. Fig. S3 shows the replenishment kinetics of the monocyte precursors and monocytes found in the BM after DT treatment of CD64^dtr mice. Fig. S4 shows the lack of major impact of DT treatment on cDCs found in the skin of CD64^dtr mice. Table S1, a separate Excel file, shows the exhaustive lists of genes responsible for the IPA enrichments shown in Fig. 7 B.

We thank M.M. Haniffa, C. Wohn, M. Guilliams, and N. Malissen for discussions; A. Adda-Benatia and F. Danjan for construction of the mice; C. Reis e Sousa for the Braf^V600E melanoma; F. Devaux, L. Chasson, and the CIML flow cytometry core facility for technical help. Microarray experiments were performed at the Plateforme GenomEast (Strasbourg, France; http://www.igbmc.fr/technologies/5/team/54/).

This work was supported by Centre National de la Recherche Scientifique, Institut National de la Santé et de la Recherche Medicale, PHENOMIN, European Research Council (FP7/2007–2013 grant 322465 to B. Malissen), Marie Skłodowska-Curie Actions (PIOF-GA-2013-625328-MeTaPATH to S. Henri), ERA-NET Infect-ERA (ABIR: ANR-13-IFEC-0005-03 to B. Malissen), the DCBIOL Labex (ANR-11-LABEX-0043, grant ANR-10-IDEX-0001-02 PSL), and the A*MIDEX project (ANR-11-IDEX-0001-02). A. Baranska is supported by a fellowship from the ABIR project. We acknowledge PICT-IBiSA, a member of the France-BioImaging national research infrastructure, supported by the CelTisPhyBio LabEx (ANR-10-LBX-0038) as part of the IDEX PSL (ANR-10-IDEX-0001-02 PSL).

The authors declare no competing financial interests.

Author contributions: B. Malissen, S. Henri, and A. Baranska conceived the project. B. Malissen and F. Fiore designed the Fcgr1-IRES-EGFP-hDTR mice. A. Baranska and S. Henri did the experiments with the technical assistance of C. Malosse and O. Voluzan. M. Malissen contributed data in Fig. 2. A. Shawket, T.-P. Vu Manh, and M. Dalod performed microarray analysis. M. Jouve and P. Benaroch performed electron microscopy. M. Baratin and M. Bajénoff performed the fluorescent microscopy analysis.