Thymic macrophages consist of two populations with distinct localization and origin

Curation statements for this article:
  • Curated by eLife

    eLife logo

    Evaluation Summary:

    The authors comprehensively dissected the ontogeny and characteristics of thymic macrophages. These findings are helpful for better understanding of the function of macrophages in thymic tissue environment.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #3 agreed to share their name with the authors.)

This article has been Reviewed by the following groups

Read the full article See related articles

Abstract

Tissue-resident macrophages are essential to protect from pathogen invasion and maintain organ homeostasis. The ability of thymic macrophages to engulf apoptotic thymocytes is well appreciated, but little is known about their ontogeny, maintenance, and diversity. Here, we characterized the surface phenotype and transcriptional profile of these cells and defined their expression signature. Thymic macrophages were most closely related to spleen red pulp macrophages and Kupffer cells and shared the expression of the transcription factor (TF) SpiC with these cells. Single-cell RNA sequencing (scRNA-Seq) showed that the macrophages in the adult thymus are composed of two populations distinguished by the expression of Timd4 and Cx3cr1 . Remarkably, Timd4 + cells were located in the cortex, while Cx3cr1 + macrophages were restricted to the medulla and the cortico-medullary junction. Using shield chimeras, transplantation of embryonic thymuses, and genetic fate mapping, we found that the two populations have distinct origins. Timd4 + thymic macrophages are of embryonic origin, while Cx3cr1 + macrophages are derived from adult hematopoietic stem cells. Aging has a profound effect on the macrophages in the thymus. Timd4 + cells underwent gradual attrition, while Cx3cr1 + cells slowly accumulated with age and, in older mice, were the dominant macrophage population in the thymus. Altogether, our work defines the phenotype, origin, and diversity of thymic macrophages.

Article activity feed

  1. Author Response

    Reviewer #1 (Public Review):

    Several questions have remained regarding the characteristics of these cells:

    1. Based on the transcriptome data in Figure 2, the authors inferred that thymic macrophages are "specialized in lysosome degradation of phagocytosed material and antigen presentation" yet did not show functional data to support these claims. Functional assays such as phagocytosis and antigen presentation are desirable, especially in comparison to other well characterized macrophage populations.

    We agree with the reviewer that additional functional characterization of thymic macrophages will strengthen the conclusions of our manuscript. We have performed antigen presentation assay and in vitro phagocytosis assay to functionally characterize the thymic macrophages. Indeed, thymic macrophages seem to be quite good antigen presenting cells – not as good as thymic DCs, but much better than peritoneal macrophages. This is documented in Fig. 3A and B. They were also good phagocytes both in vitro and in vivo as demonstrated in Fig. 3C-G. Surprisingly, peritoneal macrophages were better in the in vitro phagocytosis assay. We attribute this result to thymic macrophages’ poor survival during the sorting and in vitro culture.

    1. Do transcriptomes of CX3CR1+ thymic macrophages in old mice significantly differ from those of young mice?

    This is a very interesting question that we plan to explore in the future, but we feel it is beyond the scope of the current manuscript.

    1. It would be helpful to better graphically show the compositions (both cell number and cell ratio) of thymic macrophage subsets (TIM4+, CX3CR1+, and others) in mice at different ages (1 week, 6 weeks, and 4 months old). It is not straightforward to deduce all the information based on the current data presentation.

    We thank the reviewer for the suggestion! Plotting the cell numbers did reveal a peak in young age and then significant decline in the number of Tim4+ cells and a trend for accumulation of Tim4+ cells with age. Unfortunately, older mice show great variability in thymus size, which prevented the Tim4- result from being statistically significant. We have added these data to Fig. 8F.

    1. The description of the gating strategy of thymic macrophages for Figure 1 is quite verbose. Adding a step-wise gating strategy of thymic macrophages as a figure panel would be helpful for readers to follow the experimental details.

    We thank the reviewer for the suggestion. The description of the gating strategy has been stripped to 2 panels that capture its essence (Fig. 1B).

    Reviewer #2 (Public Review):

    This work provides by far the most thorough characterization of thymic macrophages. The authors used bulk RNA-seq, single-cell seq and fate mapping animal models to demonstrate the phenotype, origin and diversity of thymic macrophages. Overall the manuscript is well written and the conclusions of the paper are mostly well supported by data.

    Some aspects of data acquisition and data analysis need to be clarified.

    1. the authors should state what does row min row max in figure2 b,d refer to. is this expression value on log scale? In figure 2d, the authors compared their own RNAseq data with ImmGen seq data, what kind of normalization did the authors apply?

    We appologize for not making this clear. The values in Fig. 2b and d (current Fig. 2A and C) are expression values on log scale. We have included this information in the figure.

    Our data is part of the IMMGEN dataset. We sorted the cells and sent them to the US for RNA sequencing. That is why we referred to it as “our” data. However, to avoid confusion we changed the wording to clearly reflect that the data are from IMMGEN.

    2)The authors used immunofluorescent to identify the localization of two populations of macrophages, where they used merTK staining to indicate all macrophages. However, MerTK expression may not restrict to immune cells. The authors are encouraged to confirm that MerTK only labels macrophages in thymus by co-staining with F4/80 or CD45. Tim4 can also be used in immunofluorescence.

    We agree that staining with additional macrophage markers will strengthen our conclusions about ThyMacs localization. We have performed staining with CD64 together with MerTK or Tim4. CD64 and MerTK almost completely overlapped and so did CD64 and Tim4 in the cortex. We could not stain MerTK and Tim4 together because the antibodies are raised in the same species (rat). Additional evidence for the specificity of these markers for thymic macrophages comes from Fig. 3E and F showing the high degree of co-localization of apoptotic cells (TUNEL+) with MerTK or Tim4. Finally, Fig. 4 figure supplement 1 also clearly shows the distribution of TIM4 and CD64 in the whole thymus.

    1. The data of Cx3cr1+ cells accumulation with age in thymus is very interesting, and as the author has discussed, might indicate their contribution to thymus involution. However, the authors only showed change of percentage. As the total macrophages numbers decreased with age, it is not clear whether these cells actually "accumulate" with age. It will help us to assess if this increased percentage of Cx3Cr1+ cells is an actual increase of "influx" or due to the decrease of the self-maintain Tim4+ macrophage subsets.

    The reviewer is raising a very important point. As the changes in the Tim4+ and Tim4- thymic macrophages proportions with age occur at the background of thymic involution, it is difficult to judge whether Tim4+ cells self-maintain and whether Tim4- cells accumulate. Plotting the cell numbers revealed a peak in young age and then significant decline in the number of Tim4+ cells and a trend for accumulation of Tim4+ cells with age. Unfortunately, older mice show great variability in thymus size, which prevented the Tim4- result from being statistically significant. We have added these data to Fig. 8F.

    Reviewer #3 (Public Review):

    This study by Zhou et al. focuses on thymic macrophages and shows that two populations can be distinguished with different identities, localization and origin. Authors use several murine reporter and fate-mapping models, coupled with flow cytometry and transcriptomics approach to support their claims.

    Overall, the question tackled by this study is interesting, thymic macrophages having a bit being forgotten in the last decade which has seen many studies similar to the one presented here in other organs. So, the stated aim to closing this gap is relevant. But the actual version of the study suffers from many defects, more or less severe, which affect the clarity and the persuasiveness of it.

    • About the plan, authors study the origin of the thymic population and provide data in fig 2, 3 & 4 assuming that thymic macs form a homogeneous population. But from fig 5, they distinguish 2 populations and study them separately. So the end of the paper renders obsolete the beginning, that asks for a revision of the whole plan.

    We agree with the reviewer that there is more than one way to tell this story and we have been agonizing over our plan. However, we respectfully disagree that the beginning of the paper is made obsolete by the ending for several reasons:

    1. The initial figures in our manuscript contain very fundamental characterizaition of ThyMacs. Just as the revelation of a heterogeneity in liver macrophages or lung macrophages (ref) does not render all prior research on these cells obsolete, the initial figures in our manuscript are an essential part of the story. Such data are available for all other studied tissue resident macrophage populations. Removing them will be a disservice to the community.

    2. Another reviewer asked for deeper characterization of ThyMacs based on the data in Fig. 2. Accommodating this request will be very difficult if we remove this part.

    Nevertheless, we agree that ThyMacs heterogeneity is the central claim of the manuscript and should be introduced earlier. Now, the original figure 5 (current Fig. 4) that described the heterogeneity has been moved before the original figures 3 and 4 (current Fig. 5 and 6). Additional analyses distinguishing Tim4+ and Tim4- ThyMacs has been incorporated in current Fig. 5 and 6.

    • The figure 1 is not very clear. The backgating should be added in 1a. Or why not using the color map axis mode from FlowJo to show 3 parameters at a glance? The gating strategy should be more clearly displayed on the figure. On fig 1S3, there are clearly 2 pops in the CX3CR1-GFP mice. Why not starting from this to introduce the two populations?

    We thank the reviewer for the suggestion. We have included a color map axis to show MerTK, CD64, and F4/80 in one plot. The description of the gating strategy has been stripped to 2 panels that capture its essence. \We agree that there are several indications for heterogeneity among thymic macrophages, starting with Fig. 1E – the expression of Tim4, and Fig S4c – the expression of CX3CR1-GFP. We have added extra text at the beginning of the paragraph describing current Fig. 4 to point out these facts.

    • The figure 2 could be revised also. First, the panel 2a is useless and should be removed. A PC analysis of all the macs would be more useful here. Also, the color code used for the genes is confusing. Why genes up in ThyMacs are red in 2b but only half of them in 2d? Info can be found in the legend but it should be more clear on a graphical point of view.

    We have revised Fig. 2 according to the reviewer’s suggestions. The PCA analysis is consistent with the hierarchical clustering and shows that splenic and liver macrophages are most closesly related to ThyMacs. We agree that the presence of red in both heatmaps is confusing and we have changed the color code – color was removed from current Fig. 2A but retained in Fig. 2C.

    • For figure 3, what is the timepoint of the panel 3b? Here, authors should show microglia and ThyMacs for both timepoints and conclude based on the comparison. If ThyMacs are as stable as the microglia, no replacement. If not, replacement. For the panel 3f, n=3 is too low to be convinced notably with the standard variation here. And displaying the dot plot with 11% of blood mono from donor while the median being around 20 is not fair, authors should present the most representative plot. For the panel 3h, there are more GFP (in term of MFI) for TEC and ThyMacs than for total cells. How is it possible? TECs and ThyMacs should be in the total cells? Or the gating is not clear enough?

    We thank the reviewer for pointing our omissions. Fig. 3b (current Fig. 5B) is from E19.5 and we have added this information to the figure. We also agree that in Fig. 3f (current Fig. 5F) the sample number is too small and the variation too large to make solid conclusions. That is why we have repeated the partial chimeras experiment trying to irradiate as much as possible of the mice without affecting the thymus. We have substituted the data in the Fig. 3e and 3f with the new data. For Fig. 3h, we appologize for not labeling the data clearly. The panels labeled “single, live cells” should be labeled as “thymocytes” as they were obtained without enzymatic digestion that is essential for both TECs and ThyMacs. However, we found an important caveat in the thymus transplant experiment. It appeared that some of the thymus macrophages were GFP positive not because they express GFP but because they have engulfed GFP+ cells. As a result our experiments with embryonic GFP+ thymus transplants overestimate the percentage of donor-derived ThyMacs (all of them were GFP+). We have repeated the thymus transplantation experiments with congenically marked thymuses (CD45.2 donor and CD45.1 host). While this set up did not allow us to use the thymic epithelial cells as positive control because they are CD45-, we did identify host-derived ThyMacs, consistent with Tim4- cells originating from adult HSCs. Thus, we have replaced the previous data in Fig. 3H and 3I with current figures 5H and 5I.

    • For figure 4, the EdU staining (4e) is not convincing at all. The signal is very low (as compared to 4c for example.

    We agree that signal after 21d chase is a lot weaker than after 2 h (Fig. 4c) or 21d (Fig. 4e) of EdU pulse. The reason we decided to keep this data is that: 1) the thymocytes also have much lower EdU staining after 21d chase compared to 2h and 21d of EdU pulse; 2) The results from EdU staining are very consistent with the data from Ki67 staining, cell cycle analysis, and scRNA-Seq revealing a small population (~5%) of cycling ThyMacs.

    • For figure 7, the interpretation of the data and the way to present them are not clear. Authors use an inducible fate-mapping model. The fact that Tim4- loose their signal with time argue for a replacement by non-labelled cells (blood monocytes) whereas Tim4+ ones are stable meaning they self-maintain. It is what authors claim. But how it fits with previous data where they say that Tim4+ derived form CX3CR1+? The explanation that is a bit subtended here but not enough clearly shown is that CX3CR1+ give rise to Tim4+ during embryonic development but is stops after, Tim4 self-renew independently, and CX3CR1+ are slowly replaced by monocytes. As this is the central claim of the paper, it should be most clearly reported and for this, a substantial change of the whole plan is required.

    We thank the reviewer for pointing out the need for better explanation. The maintenance of the different populations of ThyMacs is indeed complex and proceeds in different ways in the different periods of life. We have added some extra data to Fig. 7 (current Fig. 8) that we hope will add some clarity to the maintenance of thymic macrophages with age. The new Fig. 8F shows the dynamics of the cell numbers of Tim4+ and Tim4- macrophages with age. Tim4+ cells reach a peak in young mice and decline significantly as mice age. So, we do not think that they are self-maintaining but instead, undergo slow attrition with very limited replacement. These results are consistent with Fig. 6I showing low levels of Mki67 in Tim4+ cells. Tim4- are a different story: they progressively accumulate with age. Although the variability in thymus size and Tim4- macrophages in very old mice is too great for the data to reach significance, the trend is clear.

    As for the dynamics of the populations in the embryonic period, we added data formally demonstrating that TIM4+CX3XR1- are derived from CX3CR1+ cells by fate mapping (Fig. 7E-G). We induced re-combination in pregnant ROSA26LSL-GFP mice pregnant from Cx3cr1CreER males at E15.5 when almost all ThyMacs are Cx3cr1+ (Fig. 7A). Just before birth, at E19.5, we could find a substantial proportion of TIM4+CX3CR1- cells among the fate mapped GFP+ macrophages, indicating that Cx3cr1+ cells, indeed, give rise to TIM4+CX3CR1- cells. As pointed out before, this pathway gets exhausted by the first week after birth – at d7 all ThyMacs are TIM4+.

  2. Evaluation Summary:

    The authors comprehensively dissected the ontogeny and characteristics of thymic macrophages. These findings are helpful for better understanding of the function of macrophages in thymic tissue environment.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #3 agreed to share their name with the authors.)

  3. Reviewer #1 (Public Review):

    Several questions have remained regarding the characteristics of these cells:

    1. Based on the transcriptome data in Figure 2, the authors inferred that thymic macrophages are "specialized in lysosome degradation of phagocytosed material and antigen presentation" yet did not show functional data to support these claims. Functional assays such as phagocytosis and antigen presentation are desirable, especially in comparison to other well characterized macrophage populations.

    2. Do transcriptomes of CX3CR1+ thymic macrophages in old mice significantly differ from those of young mice?

    3. It would be helpful to better graphically show the compositions (both cell number and cell ratio) of thymic macrophage subsets (TIM4+, CX3CR1+, and others) in mice at different ages (1 week, 6 weeks, and 4 months old). It is not straightforward to deduce all the information based on the current data presentation.

    4. The description of the gating strategy of thymic macrophages for Figure 1 is quite verbose. Adding a step-wise gating strategy of thymic macrophages as a figure panel would be helpful for readers to follow the experimental details.

  4. Reviewer #2 (Public Review):

    This work provides by far the most thorough characterization of thymic macrophages. The authors used bulk RNA-seq, single-cell seq and fate mapping animal models to demonstrate the phenotype, origin and diversity of thymic macrophages. Overall the manuscript is well written and the conclusions of the paper are mostly well supported by data.

    Some aspects of data acquisition and data analysis need to be clarified.

    1. The authors should state what does row min row max in figure2 b,d refer to. is this expression value on log scale? In figure 2d, the authors compared their own RNAseq data with ImmGen seq data, what kind of normalization did the authors apply?

    2. The authors used immunofluorescent to identify the localization of two populations of macrophages, where they used merTK staining to indicate all macrophages. However, MerTK expression may not restrict to immune cells. The authors are encouraged to confirm that MerTK only labels macrophages in thymus by co-staining with F4/80 or CD45. Tim4 can also be used in immunofluorescence.

    3. The data of Cx3cr1+ cells accumulation with age in thymus is very interesting, and as the author has discussed, might indicate their contribution to thymus involution. However, the authors only showed change of percentage. As the total macrophages numbers decreased with age, it is not clear whether these cells actually "accumulate" with age. It will help us to assess if this increased percentage of Cx3Cr1+ cells is an actual increase of "influx" or due to the decrease of the self-maintain Tim4+ macrophage subsets.

  5. Reviewer #3 (Public Review):

    This study by Zhou et al. focuses on thymic macrophages and shows that two populations can be distinguished with different identities, localization and origin. Authors use several murine reporter and fate-mapping models, coupled with flow cytometry and transcriptomics approach to support their claims.

    Overall, the question tackled by this study is interesting, thymic macrophages having a bit being forgotten in the last decade which has seen many studies similar to the one presented here in other organs. So, the stated aim to closing this gap is relevant. But the actual version of the study suffers from many defects, more or less severe, which affect the clarity and the persuasiveness of it.

    - About the plan, authors study the origin of the thymic population and provide data in fig 2, 3 & 4 assuming that thymic macs form a homogeneous population. But from fig 5, they distinguish 2 populations and study them separately. So the end of the paper renders obsolete the beginning, that asks for a revision of the whole plan.

    - The figure 1 is not very clear. The backgating should be added in 1a. Or why not using the color map axis mode from FlowJo to show 3 parameters at a glance? The gating strategy should be more clearly displayed on the figure. On fig 1S3, there are clearly 2 pops in the CX3CR1-GFP mice. Why not starting from this to introduce the two populations?

    - The figure 2 could be revised also. First, the panel 2a is useless and should be removed. A PC analysis of all the macs would be more useful here. Also, the color code used for the genes is confusing. Why genes up in ThyMacs are red in 2b but only half of them in 2d? Info can be found in the legend but it should be more clear on a graphical point of view.

    - For figure 3, what is the timepoint of the panel 3b? Here, authors should show microglia and ThyMacs for both timepoints and conclude based on the comparison. If ThyMacs are as stable as the microglia, no replacement. If not, replacement. For the panel 3f, n=3 is too low to be convinced notably with the standard variation here. And displaying the dot plot with 11% of blood mono from donor while the median being around 20 is not fair, authors should present the most representative plot. For the panel 3h, there are more GFP (in term of MFI) for TEC and ThyMacs than for total cells. How is it possible? TECs and ThyMacs should be in the total cells? Or the gating is not clear enough?

    - For figure 4, the EdU staining (4e) is not convincing at all. The signal is very low (as compared to 4c for example.

    - For figure 7, the interpretation of the data and the way to present them are not clear. Authors use an inducible fate-mapping model. The fact that Tim4- loose their signal with time argue for a replacement by non-labelled cells (blood monocytes) whereas Tim4+ ones are stable meaning they self-maintain. It is what authors claim. But how it fits with previous data where they say that Tim4+ derived form CX3CR1+? The explanation that is a bit subtended here but not enough clearly shown is that CX3CR1+ give rise to Tim4+ during embryonic development but is stops after, Tim4 self-renew independently, and CX3CR1+ are slowly replaced by monocytes. As this is the central claim of the paper, it should be most clearly reported and for this, a substantial change of the whole plan is required.