The role of recruitment versus training in influenza-induced lasting changes to alveolar macrophage function

In their recent publication, Wang et al. propose that influenza infection in the lungs confers resistance to metastatic tumor growth due to trained immunity in the alveolar macrophage (AM) compartment, underpinned by enhanced pro-inflammatory cytokine production, augmented phagocytosis and metabolic rewiring¹. Although these findings are broadly in line with prior work demonstrating that viral infections can cause long-lasting changes in the AM compartment, affecting subsequent challenges^2–5, many of these earlier studies attributed functional changes to monocyte-derived AMs emerging during infection^2–4 rather than training of pre-existing AMs. In light of this discrepancy, we believe that there is insufficient consideration of AM origin in the study by Wang et al. to formally claim that the mechanism of tumor protection is trained immunity. We believe that strategies chosen for cellular identification and purification as well as data interpretation do not fully take into account the importance of recruited monocyte-derived AMs, which share functions with trained AMs, and may therefore be mistaken as such.

‘Trained immunity’ has been described as an adaptation of innate immune cells in which exposure to pathogens or their pathogen-associated molecular patterns causes functional changes in these cells (and potentially their progeny), which result in an enhanced response to secondary challenge⁶. Key to the concept of trained immunity implicated in this study is the idea of functional change within the same cellular population. Ensuring that the trained population does not change in composition because of the initial insult is therefore imperative. At steady state, in mouse models, the majority of resident AMs derive from fetal monocyte precursors that colonize the lungs in the first week of life (FeMoAMs)^7,8. However, following inflammatory or infectious events in the lungs, resident FeMoAMs can be depleted and recruited bone marrow monocytes can differentiate into AMs (BMoAMs), thereby contributing to the AM pool^2–4,9 (Fig. 1a).

Fig. 1 |. Recruitment versus training of the alveolar macrophage compartment.

a, One month after influenza infection, fetal monocyte-derived AMs (FeMoAMs, blue) as well as bone marrow monocyte-derived AMs (BMoAMs, red) developing during infection are present and might appear similar based on expression of general macrophage markers, such as CD64, MerTK and CD11c. b, Apparent training of the AM compartment. Using CD11c-positive selection for functional analysis, the AM compartment might appear to be trained due to elevated immunoreactivity contributed by BMoAMs. c, Detection of immunoreactive BMoAMs in functional assays. Using Siglec-F expression, congenic markers in bone marrow chimeric mice or lineage tracing approaches, FeMoAMs and BMoAMs can be functionally distinguished. Created with https://biorender.com.

Previous work has shown that viral infection of sufficient severity can result in the appearance of BMoAMs with enhanced pro-inflammatory cytokine production upon stimulus (immunoreactivity), and that these cells are protective upon heterologous bacterial challenge but exacerbate severity of viral infection^2,3. These observations have been made using different influenza A virus (IAV) strains and different routes of infection^2,3, including those used in this study¹. The enhanced reactivity of BMoAMs has been considered a function of so-called ‘monocyte legacy’, by which BMoAMs retain some epigenetic and transcriptional features of their monocyte progenitors¹⁰. Training at the level of myeloid progenitors might also contribute to the BMoAM phenotype¹¹; however, in the context of influenza infection, the time window for recruitment of monocytes that contribute to the BMoAM pool is 3–7 d post-infection (d.p.i.)³, which leaves little time for such progenitors to receive and integrate infection-derived signals and subsequently differentiate into monocytes available for recruitment. Earlier research by Yao and colleagues, on the other hand, shows that FeMoAMs can be trained to have high levels of pro-inflammatory cytokine production in response to adenoviral infection⁵. Given these contrasting observations, it is paramount to ensure that FeMoAMs and recruited BMoAMs are properly identified and separated for functional analyses to avoid confusion between training of FeMoAMs and the appearance of immunoreactive BMoAMs developing from recruited monocytes during inflammation.

In Wang et al.’s study¹, most AM analyses, including RNA sequencing, assay for transposase-accessible chromatin with sequencing (ATAC-seq), adoptive cellular transfer and cell culture experiments, were performed on AMs isolated from bronchoalveolar lavage (BAL) cells, which were then positively selected only by expression of CD11c. However, CD11c is expressed by BMoAMs and FeMoAMs 1 month after infection with IAV strains, including PR8 used here (Fig. 1b)^2,3. This means that any BMoAMs present would be included in the purification, resulting in a mixed population for downstream analysis. Moreover, the use of clodronate liposomes to confirm a role for trained FeMoAMs in the phenotype of tumor resistance would equally deplete BMoAMs and other transitional macrophage populations. Given that the published characteristics of BMoAMs are very similar to those ascribed here to trained AMs, including enhanced pro-inflammatory cytokine production, differential ATAC-seq landscape and metabolic rewiring^2,3, we suggest that BMoAM contamination within the isolated AM population results in the reported differences in the AM population following PR8 infection (Fig. 1b,c).

Although we believe that AM-purification approaches in this study probably included BMoAMs in experimental readouts, gating strategies to identify them may have resulted in their exclusion. By flow cytometry, FeMoAMs are defined as CD11c⁺Siglec-F⁺CD64⁺F4/80⁺MerTK⁺ (ref. 2,12,13). While BMoAMs express slightly higher levels of CD11b, major histocompatibility complex II, CD14 and CD64 (ref. 2) transiently after IAV clearance, expression of sialic acid-binding immunoglobulin-like lectin F (Siglec-F) remains low (Siglec-F^lo) both at the level of RNA and protein after IAV infection or bleomycin-induced fibrosis, which separates them from Siglec-F^hi FeMoAMs by flow cytometry with most fluorochromes in conditions in which staining and voltages are not oversaturating^2,14 (Fig. 1b,c). Cell-transfer experiments have confirmed the stable Siglec-F phenotype of these two populations². In Wang et al.’s study¹ and prior studies⁵, the authors separate interstitial macrophages (IMs) from AMs using expression of Siglec-F, where they define AMs as Ly6C⁻Siglec-F^hi and IMs as Ly6C⁻Siglec-F^lo/− (Extended Data Fig. 1a in ref. 1). Given the relatively low expression of Siglec-F in BMoAMs, this gating strategy risks that BMoAMs are excluded from the authors’ AM gate just because they resemble IMs.

Indeed, uniform manifold approximation and projection (UMAP) plots of flow cytometry data showing the kinetic analysis of cell clusters in the BAL after infection with IAV (Extended Data Fig. 1b in ref. 1) provide some evidence of BMoAM exclusion. As reported previously, the disappearance of the AM population between 4 and 10 d.p.i. and its subsequent replenishment can be seen^2,3. Concomitantly with replenishment, a population defined by the authors as IMs appears and subsequently merges with AMs at the same time point that the AM population begins to recover. By contrast, this population was previously identified as BMoAMs based on single-cell RNA sequencing combined with trajectory analysis of BAL and lungs after IAV infection². Irrespective of the confusion created by calling these cells IMs when they are in fact recovered from the non-leaky bronchoalveolar space 30 d.p.i. (Extended Data Fig. 1c in ref. 1), the kinetics and inferred Siglec-F phenotype of this population suggest that these cells are indeed BMoAMs. Furthermore, the application of this gating strategy in parabiosis experiments (Fig. 4i–m in ref. 1) might be misleading: the systematic exclusion of recruited Siglec-F^lo BMoAMs from the AM gate will underestimate the degree of AM chimerism after infection.

Finally, a standard model for exploring an involvement of monocyte-derived cells is to use Ccr2^−/− mice, mostly devoid of blood monocytes. Although the authors present cell-intrinsic evidence that AMs derived from PR8-experienced Ccr2^−/− mice have similar changes in pro-inflammatory cytokine production as those in wild-type AMs (Fig. 4e in ref. 1), they did not perform tumor-growth experiments in these mice. Of note, Ccr2^−/− mice are not completely devoid of blood monocytes¹⁵, and, thus, even a small contribution of BMoAMs might suffice to increase the immunoreactivity of the total AM population. Alternatively, FeMoAMs might be more plastic in Ccr2^−/− mice, when monocytes are unavailable, although this would not reflect the context of wild-type mice.

We believe that the limitations of this study are symptomatic of a potentially more generalized confusion in the field caused partially by loose definitions of interstitial and alveolar macrophages¹². The promiscuity of IM surface markers on monocyte-derived cells during and after inflammatory processes suggests that their application has weaknesses¹². Therefore, localization may be more appropriate for the distinction of these subsets, with AMs residing in the alveolar space and IMs in the interstitium. Thus, at time points at which the lung barrier is intact, BAL should contain AMs rather than IMs. Furthermore, it is paramount to identify and distinguish notions and markers of activation state, lineage and cellular origin, supported by proper fate mapping. Siglec-F is a case in point as it is upregulated slowly on BMoAMs recruited into the infected lung¹⁴ but remains expressed to a lesser extent on these cells than on their embryonically derived counterparts². Thus, Siglec-F is not a lineage marker but can be used, once confirmed by fate mapping, to track origin for extended periods of time².

In summary, Wang et al.¹ add to the pool of evidence for a new steady state of lung immunity after an infection or similar events. The concomitant change in AM populations and functionality has marked capacity to influence the outcome of an increasing variety of subsequent pathologies, whether positively or negatively. However, in investigating these questions, it is crucial to distinguish between the notions of cellular recruitment and training with appropriate rigor, as these processes are probably driven by different biological mechanisms, which is of utmost importance for potential therapeutic intervention.