Mos and their derivatives play a critical role in clearing pneumonia-related pathogens. First, they can directly eliminate pathogens through phagocytosis. Phagocytes recognize phagocytic targets through surface receptor binding to pathogen-associated molecular patterns and opsonins (antibodies and complement). Upon internalization, the targets are engulfed in phagosomes, which further fuse with lysosomes to generate phagolysosomes. Within these compartments, pathogens are killed or degraded under the combined actions of an acidic environment, hydrolases and reactive oxygen species (ROS) (69,70). In addition, Mos and their derived cells can generate large amounts of ROS through NADPH oxidase and mitochondrial pathways, and produce reactive nitrogen species (RNS) via inducible nitric oxide synthase, thereby further eliminating pathogens by damaging membrane lipids, proteins and nucleic acids (71-73). Notably, ROS and RNS not only directly kill pathogens but also act as key signaling molecules that regulate inflammation, phagocytosis and cellular phenotypic transitions, thereby synergistically promoting the clearance of pneumonia-related pathogens (71,72). Mo-derived cell types differ in their pathogen-clearing capacity. In vitro, MDMs exhibited the strongest phagocytic activity against Staphylococcus aureus and Escherichia Coli (E. coli), followed by cMo and immature MoDCs, whereas mature MoDCs displayed the lowest activity. Similarly, cMos and MDMs exhibited greater killing of internalized bacteria than immature and mature MoDCs, suggesting that cMos and MDMs are specialized for pathogen clearance, while MoDCs are mainly involved in antigen presentation and adaptive immune activation (74).

Importantly, beyond direct pathogen elimination through phagocytosis and the induction of oxidative/nitrosative stress, Mos and their derivative populations contribute to host defense by orchestrating innate and adaptive immune responses, thereby establishing an integrated immune clearance network (Table I; Fig. 2). Mechanistically, these cells regulate the recruitment, activation and functional polarization of other immune cell subsets through the secretion of immunomodulatory mediators, such as cytokines and chemokines, as well as through direct cell-cell communication. In addition, Mos and MoDCs can process pathogen-derived antigens and present peptide-major histocompatibility complex (MHC) complexes to T cells, thereby initiating and shaping adaptive immune responses. Collectively, Mos and their progeny serve as central coordinators of immune-mediated pathogen clearance.

Although the role of Mos in pathogen clearance is well recognized, accumulating evidence indicates that they also contribute to mediating pathogen immune evasion (Fig. 2). Recently, Yu et al (75) proposed that excessive Mo activation in the early stages of infection mediates the immune escape of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Researchers divided patients in the silent SARS-CoV-2 infection stage of COVID-19 into an asymptomatic group (positive viral test, with no symptoms or lung imaging abnormalities during observation) and a presymptomatic group (positive viral test without initial symptoms; follow-up imaging showed progressive pneumonia, confirming early-stage COVID-19). It was found that, compared with the asymptomatic group, the presymptomatic group showed an increased proportion of CD107a^hi cMos, accompanied by a marked decrease in the frequencies of CD107a^lo cMos, intMos and ncMos. Additionally, a reduced frequency of CD62L^hi CD8⁺ T_naïve cells and an increased frequency of immunosuppressive CD4⁺ NKT cells were observed. These findings suggest that cMo overactivation is accompanied by the disruption of normal differentiation processes and early immune exhaustion of T cells. Plasma protein profiling revealed that the level of stanniocalcin-1 (STC1) (an inhibitor of Mo activation) was markedly reduced in the presymptomatic group. Subsequent in vitro experiments demonstrated that treatment with recombinant human STC1 restored normal Mo differentiation and maintained the frequency of CD8⁺ Tnaïve cells, further supporting a causal relationship between abnormal Mo activation and lymphocyte exhaustion (75). In addition, the activation of triggering receptor expressed on myeloid cells 2 (TREM2) on the surface of MDMs during pneumonia is also considered to mediate pathogen immune evasion (76,77). In individuals with tuberculosis, glycosylated mycolic acid (MA)-containing lipids can induce Mφs to produce TNF and NO through the Mincle/FcRγ/CARD9 axis, accompanied by the CCL2-mediated recruitment of inducible nitric oxide synthase (iNOS)-positive M1 Mφs (77). By contrast, non-glycosylated MA-containing lipids only promote the release of CCL2 under TREM2/DAP12-dependent conditions and recruit iNOS-negative permissive Mφs that facilitate Mycobacterium tuberculosis (M. tuberculosis) survival and persistence. In addition, TREM2-DAP12 signaling selectively antagonizes the Mincle-FcRγ-CARD9-mediated anti-mycobacterial immune response. Trem2 knockout can enhance Mincle-induced Mφ activation and the corresponding inflammatory response, and accelerate the clearance of M. tuberculosis (77). Moreover, other evidence suggests that TREM2-related signaling may modulate pneumonia pathogen clearance by regulating MDM-mediated phagocytosis and cytokine production (78,79), although the specific mechanisms remain unclear.

A successful host response to infection is invariably accompanied by an inflammatory process. Under ideal circumstances, Mos exert appropriate immune defense functions against invading pathogens, enabling the host to effectively control the infection (80). However, when pathogens with high virulence, a high burden or immune evasion ability persist in the lungs, the host initiates an inflammatory cascade characterized by the continuous and excessive recruitment and activation of Mos. This dysregulated response drives excessive inflammation, resulting in immune-mediated lung injury that may disseminate systemically and ultimately exacerbate clinical outcomes. In fact, these excessive Mo-mediated inflammatory responses often fail to achieve effective pathogen clearance (16,81,82).

The CCL2/CCR2 axis is the principal pathway mediating the recruitment of cMos. CCL2, a Mo chemotactic factor, is mainly secreted by various immune cells, including resident AMs (ResAMs), as well as by lung epithelial cells during pneumonia. The production of CCL2 is regulated by multiple factors, including pro-inflammatory cytokines and interferon (IFN) signaling pathways (82-84). Meanwhile, recruited cMos can further amplify their own recruitment through the activation of the type I IFN signaling pathway (85). Notably, as observed in coronavirus infections, impaired or delayed type I IFN responses may increase viral replication, which in turn activates NF-κB, increases CCL2 production and ultimately promotes excessive cMo recruitment (86-88). Other pathways, including the CCL3,4,5/CCR5 axis (89) and complement activation [for example, C3a induction in influenza A virus (IAV) pneumonia (90)], also contribute to excessive cMo infiltration.

Once recruited, cMos and MDMs, which are influenced by pathogens, immune cells and immune mediators (17,91), frequently adopt an excessively pro-inflammatory phenotype, thereby creating a highly inflammatory pulmonary microenvironment and driving immune injury through increased production of inflammatory cytokines, chemokines and other mediators. Notably, compared with ResAMs, recruited BMo-AMs exhibit stronger inflammatory signaling, increased glycolysis, and increased production of IL-1β and IL-6 (92). Recruited IMs likewise exhibit pro-inflammatory transcriptional profiles (76,93). The acquisition of a more pro-inflammatory phenotype by MDMs may be attributed to their retention of the chromatin landscape and transcriptomic features of precursor Mos and/or to the differentiation of their precursor Mos within an inflammatory environment. Additionally, the evidence from a study of COVID-19 indicates that intMos and/or ncMos may modulate CD4⁺T cell polarization toward Th1 and Th17, which is consiered to potentially contribute to an increased severity of pneumonia (66).

Accumulating evidence indicates that severe pneumonia is frequently characterized by a lung inflammatory milieu dominated by cMos and MDMs. In patients with severe COVID-19, these cells exhibit high expression of pro-inflammatory cytokine and chemokine genes, accompanied by elevated levels of inflammatory mediators in bronchoalveolar lavage fluid (BALF) (24,94-96). A similar cMo/MDM-driven inflammatory pattern has been observed in IAV-induced pneumonia, where cMos and M1-like MDMs orchestrate the inflammatory microenvironment (97). In bacterial pneumonia, mild cases are characterized by effective humoral responses, whereas severe cases are characterized by BALF enriched with inflammatory MDMs expressing CXCL1,3,8, TNF, IL6 and IL1B, where chemokine-mediated cross-talk among these inflammatory Mφ subsets further amplifies the inflammatory cascade (98).

During the onset and progression of pneumonia, circulating Mos display complex and heterogeneous immunophenotypes. The levels of enrichment of circulating Mos with distinct phenotypes varies considerably across patient populations (23,99,100). Below, the relatively common proinflammatory circulating Mos, which can contribute to and exacerbate immunopathological injury, are discussed; however, Mos with immunosuppressive or mixed phenotypes also exist, highlighting the heterogeneity of circulating Mo populations (101). Several studies of COVID-19 have shown that circulating Mos exhibit a pro-inflammatory phenotype and constitute a major source of elevated levels of pro-inflammatory cytokines in peripheral blood (31,102-104). In patients with bacterial pneumonia, cMos are the major contributors to inflammation and cytokine production. These cells activate their intrinsic TLR4-MYD88 signaling pathway through the secretion of S100A8/A9/A12, thereby forming a positive feedback loop that aggravates immune-mediated injury (105). cMos may also amplify inflammation via ligand-receptor interactions with other inflammatory cell clusters. Moreover, a C1QA/B/C-high ncMo subset, particularly enriched in severe disease, exacerbates inflammation and immune injury through complement activation. Circulating Mos can further acquire a pro-thrombotic phenotype, thereby contributing to the formation of immunothrombosis. In patients with COVID-19, cMos exhibit increased expression of hemostasis- and platelet activation-related genes and increased Mo-platelet aggregate formation, which correlates with disease severity (106).

Finally, the mononuclear phagocyte system in the peripheral circulation and the lungs is not independent; rather, it cooperatively drives local and even systemic inflammatory responses through complex cross-talk. In patients with IAV infection, the expansion of M1-like Mos in peripheral blood leads to the secretion of large amounts of TNF-α, thereby mediating severe immune-mediated tissue injury. Concurrently, their recruitment into the lung may increase the M1/M2 Mφ ratio, further exacerbating pulmonary pathological damage (107). Additionally, in patients with COVID-19, the lung contains highly pro-inflammatory Mono_c1-CD14-CCL3 cells that, through CCR5 expression, can respond to stimuli from a variety of cells in both the lung and peripheral blood (31).

In the late stage of pneumonia, MDMs exhibiting anti-inflammatory and reparative phenotypes play a crucial role in resolving inflammation and promoting lung tissue repair. Conversely, Mos and MDMs may also contribute to the development of pneumonia-related PF (Fig. 3). This section will review the specific roles of Mos and MDMs in these two processes, as well as the complex mechanisms involved.

MDMs can mediate the resolution of inflammation and the repair of lung tissue in patients with various lung injuries, including pneumonia (28,108-110). Repair-associated MDMs may originate from either the phenotypic evolution of early injury-recruited MDMs or from distinct later waves of Mos (15,111,112). After fulfilling their reparative functions, these cells may either integrate into the lung resident tissue macrophages (RTM) pool (28) or undergo clearance (113).

Watanabe et al (111) proposed two non-mutually exclusive models by which MDMs facilitate inflammation resolution and tissue repair: The passive and active repair models. In the passive model, the recovering tissue microenvironment provides progressively stronger steady-state signals that gradually steer MDMs toward RTM-like phenotypes and functions, thereby forming a positive feedback loop that supports tissue normalization; alternatively, MDM apoptosis may facilitate RTMs restoration. In the active model, late-stage injury signals, including efferocytosis, Tregs and epithelial-derived factors, activate specific transcriptional programs in MDMs, inducing the secretion of repair-promoting mediators such as anti-inflammatory molecules, growth factors, matrix metalloproteinases, osteopontin and lipid mediators. IL-4 and IL-13 produced by Th2 cells and group 2 innate lymphoid cells (ILC2s) also serve as key drivers of anti-inflammatory and reparative MDM phenotypes (15,112,113).

The aforementioned term 'efferocytosis' refers to the recognition, phagocytosis, and clearance of dying cells and cell debris by Mφs. This process prevents the release of harmful intracellular contents, helps terminate inflammatory responses, promotes tissue repair and restores lung homeostasis. The review by Kourtzelis et al (114) provides a detailed description of the specific mechanisms underlying Mφ-mediated efferocytosis. In most cases, efferocytosis drives Mφs toward an anti-inflammatory, pro-repair phenotype, which represents a key step in re-establishing homeostasis following tissue injury (111,114). However, an in vitro study reported that efferocytosis of SARS-CoV-2-infected apoptotic cells by MDMs inhibited the acquisition of an anti-inflammatory phenotype while promoting a pro-inflammatory phenotype characterized by increased secretion of IL-6 and IL-1β. This phenomenon appears to be related to SARS-CoV-2 activity and may be specific, as infection with coxsackievirus did not induce a similar phenotype in MDMs. Moreover, this process impairs the efferocytic capacity of MDMs, which may result in the accumulation of damage-associated molecular patterns in the lung and consequently exacerbate immune-mediated injury. However, the impairment of efferocytosis does not depend on SARS-CoV-2 RNA replication (91).

In addition, a recent study identified a short-lived Ly6G-positive (a neutrophil marker) Mφ subset in the alveolar lumen surrounding the lesion area during the early recovery phase in IAV-infected mice (113). These cells are derived from recruited cMos (CD64⁻Ly6C⁺) through inflammatory Mos (CD64⁺Ly6C⁺), express high levels of SPP1 and Arg1, and exhibit robust metabolic activity, phagocytosis and efferocytosis. Ly6G⁺ Mφs can directly act on AT2 cells by releasing soluble factors, including chemokines, cytokines and osteopontin, thereby supporting AT2 cell proliferation and establishing a permissive environment for the AT2-to-AT1 transdifferentiation program, a process that depends on type II cytokine-mediated stimulation of IL-4R on Ly6G⁺ Mφs. Similar MDM populations were also observed in non-infectious lung and liver injury models and in BALF from patients with suspected pneumonia, suggesting a broader role in tissue repair after lung injury.

Several studies have reported persistent epigenetic changes in circulating Mos weeks to months after COVID-19, suggesting the establishment of innate immune memory (152-156). The reprogramming of hematopoietic stem and progenitor cells (HSPCs) in the bone marrow is considered the key to maintaining innate immune memory in circulating Mos (140-142,157-160). Notably, the work by Cheong et al (42) has received widespread attention (42,150,151). Specifically, it was demonstrated that severe COVID-19 can induce durable epigenetic and transcriptional reprogramming of HSPCs, thereby exerting prolonged effects on patients' immune function for up to 1 year. On the one hand, this reprogramming causes a long-term skewing of myelopoiesis, which is characterized by a marked and sustained increase in the frequency of GMPs; on the other hand, the epigenetic and transcriptional reprogramming of HSPCs is transmitted to descendant Mos, leading to the enrichment of epigenetic and transcriptional programs related to activation, differentiation, antigen presentation and antiviral responses, and conferring on Mos features similar to those of intMos and DCs. Mechanistically, the IL-6 signaling pathway is a key driver of the persistent phenotype of HSPCs. The data from clinical cohorts and mouse infection models consistently indicate that IL-6R blockade treatment during the acute phase can alleviate long-term myelopoiesis skewing and immune function remodeling following infection.

Consistent with the findings of Cheong et al (42), Denstaedt et al (41) recently studied preclinical sepsis models and showed that sepsis can, on the one hand, reprogram Mos (and their progenitors) into a pro-inflammatory phenotype by enriching the AP-1 motif and activating signaling pathways such as the JAK-STAT pathway; on the other hand, sepsis also reprograms the hematopoietic system by expanding GMPs in the bone marrow and consequently increases downstream NeuMo production. Upon subsequent secondary LPS challenge in the lungs, these reprogrammed Mos mediate more severe tissue damage by activating neutrophils and inducing their degranulation. Consistently, circulating Mos in patients with community-acquired pneumonia (CAP) exhibit upregulated expression of neutrophil-related genes, indicating that pneumonia similarly induces innate immune memory in Mos and their progenitors.

Current views suggest that trained immunity has dual regulatory characteristics: It can provide immune protection, yet may simultaneously drive excessive inflammation or immunopathological damage (141). Similarly, trained immunity induced by pneumonia exerts a 'double-edged sword' effect. On the one hand, in vitro experiments have shown that trained Mos can secrete large amounts of pro-inflammatory cytokines upon re-stimulation (42,155). While these cytokines are critical for early pathogen clearance, they may also exacerbate tissue pathology. Due to the lack of in vivo studies that systematically assess the overall function of trained Mos during secondary infection, whether their net effect predominantly enhances immune protection or primarily promotes immunopathology remains unclear. On the other hand, evidence from patients with COVID-19 suggests that trained Mos may represent one of the potential mechanisms underlying post-acute sequelae of SARS-CoV-2 infection (PASC) (151). PASC is characterized by persistent chronic inflammation (81,161). Notably, although not specific, pro-inflammatory Mos with phenotypes similar to the trained Mos aforementioned have been detected in the peripheral blood of patients with PASC (162,163). Moreover, Cheong et al (42) demonstrated that trained Mos can be persistently recruited to the lungs and brain in preclinical models, where they mediate immunopathology. Notably, although less well recognized, sequelae have been reported for viral pneumonias other than COVID-19 (81,161,164). Elucidating the role of trained immunity in Mos and their progenitors in the development of pneumonia-related sequelae is an important question warranting further investigation in the future.

In addition to enhancing innate responses, pneumonia can also induce a durable tolerized state in circulating Mos. During the acute phase of COVID-19 and other CAPs, peripheral blood Mos frequently exhibit altered cytokine responsiveness, reduced Mo human leukocyte antigen-DR (mHLA-DR) expression and/or impaired antimicrobial functions (32,165-171), despite the frequent coexistence of systemic inflammation (26,166-170,172,173). In fact, previous studies of the tolerant phenotype of Mos have mainly focused on sepsis (174-177), and this phenomenon has only been gradually recognized and appreciated in the context of pneumonia in recent years. Notably, two recent consecutive studies conducted by the same research group preliminarily explored the transcriptomic, epigenetic and metabolic features of tolerant Mos in the peripheral blood of patients with CAP (178,179). Importantly, this tolerant phenotype has been reported to persist for weeks to months following both mild and severe pneumonia (178,180,181), arguing against a purely transient deactivation state and instead suggesting durable immune imprinting. From a clinical perspective, the 'tolerant' state may help mitigate excessive inflammatory responses, but it also increases the risk of secondary infections. Moreover, these observations, especially the reduction in mHLA-DR (165), provide a biological rationale for dynamic immune monitoring and for considering immunomodulatory interventions in selected patients.

The precise mechanisms underlying the development of the tolerance phenotype in circulating Mos may be associated with large-scale cytokine production during pneumonia. On the one hand, according to the conjecture proposed by Joshi et al (175), pro-inflammatory cytokines drive rapid myeloid proliferation; however, this process outpaces Mo metabolic maturation and/or adequate GM-CSF stimulation, rendering newly generated Mos metabolically insufficient and therefore prone to tolerance. On the other hand, various cytokines can also directly influence the phenotype of circulating Mos (23,168,169). However, the aforementioned mechanism does not appear to provide a satisfactory explanation for the long-term persistence of Mo tolerance. Instead, this phenomenon suggests reprogramming at the HSPC level (178,181); however, the specific process remains unexplored, and cytokines may also play a key role. In vitro experiments have demonstrated that IL-6 and IL-10 drive HSPCs to differentiate into MS1 Mos, which exhibit a tolerant phenotype characterized by reduced cytokine induction and the suppression of T-cell responses (182). In addition, Roquilly et al (139) proposed a further hypothesis. Specifically, it was speculated that signal-regulatory protein α (SIRPα) functions as a pivotal sensor of inflammation and as a trigger for tolerogenic training throughout the body, and that in vitro anti-SIRPα treatment can restore the phagocytic capacity of circulating Mos.

AM remodeling represents another important form of pneumonia-induced innate immune memory. Before birth, fetal Mos colonize the alveoli and differentiate into fetal Mo-derived AMs (FeMo-AMs), which constitute the initial ResAM population. Under physiological conditions, ResAMs maintain alveolar microenvironmental homeostasis without eliciting unnecessary inflammation. In the steady state, ResAMs can be replenished through either self-renewal or the recruitment of circulating Mos. The prevailing view is that ResAMs primarily sustain their numbers through local proliferation, with only a limited contribution from external input (60,183,184); in other words, FeMo-AMs account for the majority of ResAMs that have not experienced infection or injury. However, some studies suggest that, under steady-state conditions, the proportion of BMo-AMs may gradually increase with age (38,185). Pneumonia leads to the depletion of ResAMs, and both the self-renewal of residual ResAMs and the recruitment of BMo-AMs act to replenish the depleted AM pool. Pneumonia not only remodels the compositional landscape of AMs but, more importantly, induces durable functional alterations in AMs that persist even after clinical recovery. The mechanisms underlying these persistent functional changes in AMs after pneumonia remain incompletely understood; potential contributing factors include a 'Mo legacy', the influence of inflammatory cues on Mo differentiation into BMo-AMs, the training of AMs of distinct ontogenies by the lung microenvironment, and the training of BMo-AM precursor cells, such as Mos and hematopoietic stem cells (HSCs) (Fig. 5A). Notably, although the persistent functional changes in AMs following pneumonia resemble the processes described in peripheral RTM-trained immunity, the pneumonic process involves the emergence of new cellular subsets (BMo-AMs). This result underscores the critical need to carefully distinguish between the unique functions of recruited BMo-AMs and the concept of ResAM training when investigating long-term AM functional reprogramming after pneumonia, as this distinction is critical for the development of targeted interventions (186). A number of studies have further elucidated the intricate mechanisms that may drive post-pneumonia AM remodeling (139,156,185,187), and multiple theoretical models have been proposed to describe this process (61,188), although substantial controversies and knowledge gaps remain. Table II presents a comprehensive summary of studies examining post-pneumonia AM remodeling.

Pneumonia induces extensive Mo infiltration into the alveoli, where these cells can further differentiate into BMo-AMs during inflammation, thereby mediating immune pathology while concurrently contributing to pathogen clearance (92). A portion (potentially even the majority) of recruited BMo-AMs disappears once inflammation resolves (92,183,189), whereas the remaining cells [referred to as long-lived BMo-AMs to distinguish them from transient AMs (TransAMs)] contribute to the reconstitution of the post-pneumonia AM pool. This short-lived BMo-AM subset can be referred as TransAMs, yet current research on TransAMs remains very limited (61,190). Similarly, ResAMs (of which FeMo-AMs are considered to constitute a large proportion in naïve mice) can also contribute to the post-pneumonia AM pool through self-renewal (Fig. 5B). Notably, the findings of Li et al (185) suggest that these two routes of replenishment of the AM pool occur in a temporally ordered manner. Specifically, IAV infection induces the substantial depletion of FeMo-AMs, after which the surviving FeMo-AMs promptly reconstitute the AM pool through self-renewal. Subsequently, the recruited Mos progressively differentiate into BMo-AMs, which further replenish the AM pool and ultimately supersede FeMo-AMs as the dominant component of the AM pool within weeks to months following infection clearance. By contrast, in naïve mice, bone marrow-derived Mos exhibit a weaker ability to colonize empty AM niches than fetal Mos. One possible explanation is that lung infection increases alveolar glucose levels, thereby increasing glycolysis and proliferation in BMo-AMs.

A number of studies have relied on the 'niche' model (61,187,188,190), a framework originally proposed by Guilliams et al (191,192), to explain the divergent proportions of BMo-AMs and FeMo-AMs observed after pneumonia. The Mφ niche constitutes a multifunctional microenvironment that provides structural support, trophic factors for self-renewal and tissue-specific signals that shape Mφ identity, while Mφs in turn help maintain the niche (191,192); the AM niche may comprise alveolar epithelial cells, fibroblasts, ILC2s and basophils (61,193). Guilliams et al (191) further suggested that the extent of RTM depletion, the intensity and nature of inflammation, and Mo access to the niche collectively determine the proportion of MDMs within the RTM pool during inflammatory remodeling. Accordingly, numerous scholars posit that the remodeling of the composition of the AM pool after pneumonia depends largely on the extent of ResAM depletion, namely, the severity of pneumonia (61,188,190,194).

This pioneering theory provides a reasonable explanation for the discrepancies in the results of the existing research, but several contradictory findings also merit discussion. On the one hand, in the study by Wang et al (195), although IAV infection depleted nearly 90% of ResAMs, their replenishment still relied predominantly on ResAM self-renewal, with BMo-AMs contributing only minimally. This result suggests that, in addition to the extent of ResAM depletion, other factors must be considered when determining the composition of the AM pool after pneumonia, such as differences in competitive capacity between ResAMs and recruited BMo-AMs for vacant niches. On the other hand, niche availability represents one of the key variables in the 'niche' model, which posits that once organ growth ceases, niches become fully occupied, thereby limiting the contribution of circulating precursors unless niches reopen (192). The limited niche number does not refer to a restriction in physical space but rather to the availability of trophic factors (51,191,196). Notably, in some mouse models of viral pneumonia, the number of AMs after pneumonia actually increased compared with the baseline number (185,194,197), and similar phenomena were observed even in Ccr2^−/− mice (194). The mechanisms underlying this phenomenon remain unclear; pneumonia may induce the generation of new niches (185) or could indicate the presence of more complex and as yet unexplained mechanisms within the 'niche' model.

Moreover, differences in mouse strains may be a major contributor to these contradictory results (197,198). A typical example is that Gilliaux and Desmecht (197) found that the AM pool in BALB/c mice was markedly expanded 28 days after intranasal infection with MuHV-4 compared with that in control mice, whereas no differences were observed in mice of the CD-1 and C57BL/6 strains. In addition, the Siglec-F fluorescence (Siglec F^lo AMs represent BMo-AMs, whereas Siglec F^hi AMs represent FeMo-AMs) in BALB/c MHC II^hi AMs showed a bimodal distribution, suggesting dual AM origins; in contrast, in CD-1 and C57BL/6 mice, Siglec-F exhibited a unimodal distribution and its levels remained comparable to those in control mice, indicating the self-renewal of ResAMs.

Pneumonia can also induce long-lasting alterations in AM function (Fig. 5C). Given that recruited BMo-AMs exhibit greater plasticity than ResAMs, researchers presumed that these recruited BMo-AMs play a central role in the sustained functional reprogramming of AMs following pneumonia. Several theoretical models have been proposed to delineate the potential scenarios involved in this process.

Based on the work of Aegerterg et al (194), Kulikauskaite and Wack (188) further proposed the theoretical model of the 'Mo legacy'. It was argued that BMo-AMs can, to some extent, retain specific features of the chromatin landscape and transcriptional profile of Mos, and that the pulmonary inflammatory environment may contribute to preserving the increased immunoreactivity of BMo-AMs inherited from Mos. As inflammation subsides, the pulmonary homeostatic environment influences BMo-AMs through two main pathways, cytokine signals, such as TGF-β and GM-CSF, and metabolic constraints, which together remodel their phenotype and ultimately convert them into steady-state cells resembling ResAMs that are characterized by tissue maintenance and low immunoreactivity. The authors also highlighted that the high reactivity of BMo-AMs may also arise from training during the differentiation of HSCs into BMo-AMs, or may be acquired by BMo-AMs within the inflammatory pulmonary microenvironment (141,158,199,200). By contrast, Guilliams and Svedberg (61) posited that the inflammatory differentiation trajectory of Mos, rather than cellular ontogeny, underly the sustained hyperreactivity of recruited BMo-AMs. Specifically, they infer that tissue homeostatic signals constrain the transcriptional plasticity of ResAMs, thereby preventing excessive inflammation-induced tissue damage. Consequently, inflammatory signals during pulmonary inflammation elicit only modest effects. By contrast, after recruited Mos enter the tissue, their differentiation is shaped by dual imprinting from both tissue-specific and inflammatory signals, ultimately giving rise to inflammation-imprinted resident AMs (InfResAMs) with high plasticity that retain certain pro-inflammatory features for a period following the resolution of inflammation. Over time, under the influence of homeostatic signals, InfResAMs undergo phenotypic remodeling and progressively acquire characteristics resembling ResAMs.

Together, these models suggest that recruited BMo-AMs acquire distinct functions following pneumonia but, in response to homeostatic pulmonary cues, gradually become ResAM-like and contribute to the ResAM pool, although some origin-dependent transcriptional differences persist (201,202). However, the intrinsic differences between ResAMs of distinct origins and their functions may have long been underestimated. A recent study by Li et al (185) suggested that BMo-AMs themselves indirectly tend to mediate more severe immunopathology during IAV infection than FeMo-AMs by modulating the activity of other cells within the niche. This important finding suggests that once conditions such as infection or aging expand the BMo-AM fraction within ResAMs, they may potentiate subsequent pneumonia, underscoring the need to re-evaluate the heterogeneity between BMo-AMs and FeMo-AMs and to define the specific functional roles of BMo-AMs, rather than treating them as interchangeable homeostatic populations. Additionally, although ResAMs are generally considered to exhibit limited plasticity and to be only modestly influenced by the pneumonia microenvironment (185,194), studies have reported infection-induced training effects on ResAMs (139,156,195,203), suggesting that their contribution to long-term AM reprogramming should not be overlooked. Furthermore, the preclinical evidence from pneumococcal pneumonia shows that both FeMo-AMs and recruited BMo-AMs undergo pronounced and sustained phenotypic changes, even though FeMo-AMs constitute only a minor AM subset, suggesting that the lung microenvironment is a key driver (204). However, why recruited BMo-Ams, with their high plasticity and potential 'Mo legacy', exhibit post-pneumonia phenotypes comparable to those of the less plastic FeMo-AMs remains unclear, especially given that FeMo-AMs can even acquire an increased phagocytic capacity following infection (204). Finally, one study reported that recruited IMs in models of LPS-induced lung injury initially exhibit robust inflammatory signatures, which subsequently wane as their transcriptional profile progressively converges toward that of resident IMs, while still maintaining Il1b upregulation, a characteristic potentially associated with innate immune memory (93). However, the current understanding of the precise changes in the abundance and functional properties of IMs during pneumonia progression and throughout the recovery phase remains limited (15,61,205).

In summary, pneumonia can induce sustained remodeling and functional alterations in AMs, with recruited BMo-AMs potentially playing a key role. Although this phenomenon has been extensively discussed (15,193), its precise mechanisms and potential modulatory factors remain incompletely understood. Moreover, ResAMs appear to independently contribute to AM remodeling, and the use of different experimental models and protocols may yield conflicting results. Therefore, future research should prioritize delineating the respective contributions of AMs from different ontogenies in post-pneumonia remodeling and rigorously defining the underlying mechanisms to avoid misleading researchers performing subsequent studies. Notably, the majority of the existing evidence is derived from animal experiments, in which the number and sequence of infections can be strictly controlled and defined. Furthermore, most animal models examined in previous studies were free from preceding respiratory infections and other diseases, enabling a direct observation of AM remodeling after pneumonia. However, the situation in human populations is far more complex. As highlighted in a number of reviews (188,190,193), due to factors such as age, a prior infection history and other pulmonary diseases, human AMs may have undergone long-term remodeling, potentially resulting in discrepancies between the preclinical experimental results and actual clinical conditions.

Interindividual variations in Mo phenotypes may underlie the clinical heterogeneity of pneumonia. These phenotypic variations may reflect disease endotypes or stages and are closely associated with the disease severity, treatment response and clinical outcomes (23,99,100). Given that circulating Mo phenotypes are readily detectable in peripheral blood, which can be collected minimally invasively and repeatedly for dynamic monitoring, they hold promise as valuable tools for the stratification, risk assessment and personalized treatment decision-making in patients with pneumonia (26,32,173).

Liu et al (99) identified three heterogeneous patient clusters by integrating single-cell transcriptomes of peripheral blood mononuclear cells from patients with COVID-19, primarily based on distinct inflammatory phenotypes of Mos, and proposed a 'three-stage' model. Cluster 1 is characterized by enhanced adaptive T-cell immune responses and subtle Mo inflammatory features, and predominantly comprises healthy donors and convalescent patients. Cluster 2 is characterized by a markedly increased proportion of immunosuppressive MS1-like Mos (exhibiting high expression of S100A8, S100A9, IL18 and RETN, with low expression of MHC II genes), resembling the immunosuppressive phenotype observed in patients with sepsis, with the majority of patients in the convalescent phase and a small subset in the active phase. Cluster 3 is characterized by a markedly increased proportion of hyperinflammatory Mono-CD14-CCL3 Mos and megakaryocytes displaying high expression of pro-inflammatory cytokines such as IL6, IL1B, CCL3 and TNF, which contribute to the development of a cytokine storm, and predominantly consists of patients with severe active-phase disease (99). Heterogeneous endotypes in patients with pneumonia may necessitate differentiated clinical interventions.

Furthermore, the detection of mHLA-DR offers a more convenient approach for assessing the disease state of patients with pneumonia than a single-cell transcriptomic analysis. Indeed, while circulating mHLA-DR expression levels have been well-established as a critical indicator for evaluating immunosuppression in patients with sepsis (175,176,206), their clinical significance in patients with pneumonia remains incompletely understood. In recent years, accumulating evidence from studies of COVID-19 has have shown that circulating mHLA-DR expression progressively decreases with increasing COVID-19 severity. Reduced mHLA-DR levels are markedly associated with poor pneumonia outcomes (165,167,168,207,208) and may serve as a robust biomarker for predicting disease severity and/or the mortality risk (208-211). Importantly, circulating mHLA-DR levels also facilitate the identification of pneumonia subphenotypes. Marais et al (100) reported that among adult patients with severe COVID-19, persistently high mHLA-DR expression (hyperactivated Mo/Mφ phenotype) correlated with increased mortality, whereas persistently low mHLA-DR expression was associated with secondary infections. Similarly, by leveraging differences in ferritin and mHLA-DR levels, patients with severe respiratory failure due to COVID-19 were successfully classified into two distinct subgroups: Those with Mφ activation-like syndrome (MALS) and those with immune dysregulation (23). In addition, there have been reports examining the clinical importance of circulating mHLA-DR levels in patients with non-COVID-19 pneumonias, although the available evidence remains limited and predominantly derived from studies of critically ill patients (212-216). The detailed findings of these studies are summarized in Table III.

In summary, the cellular and molecular signatures of circulating Mos may help identify clinically relevant pneumonia endotypes or subphenotypes and facilitate personalized, precision management.

Therapeutic strategies targeting Mos, including IFN-γ, GM-CSF and the inhibition of the IL-1 pathway, remain inadequately explored in patients with pneumonia. Nevertheless, several randomized controlled trials (RCTs) conducted in patients with sepsis have evaluated these immunomodulatory approaches and may provide indirect, hypothesis-generating evidence to guide therapeutic development and the design of trials of these drugs for pneumonia.

In terms of anti-inflammatory therapy, IL-1 pathway blockade therapy has garnered widespread attention. Anakinra is a recombinant IL-1 receptor antagonist that blocks IL-1 receptor type 1 signaling mediated by both IL-1α and IL-1β. In two early large RCTs of unselected patients with sepsis, IL-1 receptor antagonist did not result in an overall survival benefit (217,218). However, a post hoc reanalysis of one of these studies demonstrated that anakinra was associated with reduced 28-day mortality among patients with hepatobiliary dysfunction and disseminated intravascular coagulation, a phenotype characteristic of Mφ activation syndrome (MAS) (219). MAS is characterized by the aberrant activation of Mφs and CD8⁺ T cells, which leads to a cytokine storm and a systemic hyperinflammatory state, with IL-1β serving as a pivotal mediator of its pathogenesis (220). Notably, IL-1β-driven MALS can also be observed in patients with pneumonia, suggesting a potential therapeutic effect of anakinra (221,222). The recent phase 2a INSPIRE RCT investigated the efficacy of anakinra in patients with non-COVID-19 pneumonia (223). Specifically, presepsin was employed as a marker to indicate the initiation of the early IL-1-mediated inflammatory cascade in Mos/Mφs obtained from patients. In hospitalized patients with a qSOFA score (224)=1 and presepsin levels >350 pg/ml, anakinra administration may reduce organ dysfunction progression and 90-day mortality while accelerating discharge (223). For COVID-19, a Cochrane review of four RCTs in the study by Davidson et al (225) demonstrated no evidence of a benefit from anakinra in patients with COVID-19, whether assessed by a clinical improvement, the proportion of patients with World Health Organization clinical progression scores (226) ≥7 or all-cause mortality. However, a pooled analysis of three non-RCTs conducted by Wang et al (227) demonstrated that anakinra treatment markedly reduced mortality among patients with COVID-19 with a SOFA score ≥2.

Canakinumab is a fully human monoclonal antibody that neutralizes IL-1β. Unlike anakinra, it is primarily indicated for the management of autoinflammatory diseases, and its application in treating sepsis or pneumonia has remained limited, with only a few RCTs assessing its therapeutic efficacy against COVID-19 (228,229). A phase 3 RCT involving 454 patients showed that for those patients with severe COVID-19 with systemic hyperinflammation but not requiring invasive mechanical ventilation, canakinumab treatment did not improve the survival rate without invasive mechanical ventilation at day 29 (228).

GM-CSF and IFN-γ-based immunostimulatory therapies are considered to be capable of restoring normal immune function in Mos/Mφs from septic patients with immunoparalysis (175,206). With respect to Mo function, several RCTs have demonstrated that GM-CSF increases mHLA-DR expression in septic patients (230-233), with one study additionally reporting the restoration of the Mo capacity to secrete pro-inflammatory cytokines (230). Meanwhile, the findings from early RCTs indicate that, compared with the placebo, GM-CSF treatment markedly improves infection cure/improvement rates, shortens the duration of mechanical ventilation, improves APACHE-II scores in patients with sepsis and increases the partial pressure of arterial oxygen to fraction of inspired oxygen ratio in individuals with respiratory dysfunction (230,231,234). However, in a recent multicenter RCT involving 98 septic patients with low mHLA-DR levels, no marked differences were observed between patients receiving GM-CSF therapy and those receiving the placebo in terms of the incidence of ICU-acquired infections, the frequency of mechanical ventilation or renal replacement therapy, the duration of the ICU stay or the length of the hospital stay (233). Notably, the findings of this study were influenced by the limited sample size resulting from the early termination of the trial (233). Furthermore, the results of multiple RCTs have consistently shown that GM-CSF therapy does not reduce the risk of mortality in patients with sepsis (230,231,233,234).

A small RCT demonstrated that IFN-γ treatment markedly increases TNF-α levels and mHLA-DR expression in healthy volunteers administered E. coli endotoxin, with more pronounced therapeutic effects than GM-CSF (235). Recent RCTs findings suggest that IFN-γ therapy may provide a clinical benefit for septic patients with immunoparalysis, as described next.

Notably, two recent RCTs have investigated the clinical efficacy of precision immunotherapy by identifying septic patients with distinct immune statuses (236,237). These findings provide an important reference for the clinical application of the aforementioned Mo-targeted therapeutic strategies in the context of precision medicine for patients with pneumonia. The PROVIDE RCT stratified patients with sepsis according to their immune status into three distinct categories: MALS (ferritin level >4,420 ng/ml), immunoparalysis (mHLA-DR expression <5,000/cell) and an intermediate group. Among the 240 enrolled patients, these groups comprised 20.0, 42.9 and 37.1% of the cohort, with corresponding 28-day mortality rates of 79.1, 66.9 and 41.6%, respectively. Within the MALS subgroup, anakinra markedly increased the proportion of survivors achieving a reduced SOFA score by day 7 (42.9 vs. 10.0%); however, it did not improve 28-day survival, which was potentially attributable to the limited 7-day treatment duration (236). Additionally, the initial limitations in the diagnostic criteria for immunoparalysis resulted in only a small number of patients receiving IFN-γ therapy, thereby precluding a meaningful efficacy evaluation (236). Overall, the PROVIDE trial offers preliminary evidence supporting immune-based stratification and targeted therapy in sepsis management. A subsequent RCT (237) adopted this classification criterion and employed an extended treatment duration (15 days) to further investigate the clinical efficacy of precision immunotherapy strategies in patients with sepsis. This study found that in the precision immunotherapy group (anakinra for MALS and IFN-γ for immunoparalysis), the proportion of patients achieving a mean decrease in the SOFA score of ≥1.4 points by day 9 was markedly increased (35.1 vs. 17.9%). The treatment also promoted immune recovery and infection resolution (237). However, consistent with the findings of the PROVIDE RCT (236), no improvements in 28- or 90-day mortality were observed. Notably, anakinra was associated with a higher incidence of anemia, whereas IFN-γ was associated with more hemorrhagic events (237).

Moreover, the therapeutic interruption of defined pathogenic signaling pathways may theoretically mitigate pneumonia-associated immunopathology. For instance, preclinical studies have demonstrated that the blockade of TNF and IFN-γ attenuates the generation of pro-fibrotic MDMs, thereby alleviating PF in patients with COVID-19 (27,134). However, these interventions may also disrupt other essential pathophysiological processes and lead to unforeseen consequences. These concerns, together with other limitations inherent to preclinical models, have impeded the translation of promising experimental findings into clinical practice. Further mechanistic studies and clinical validation are needed to facilitate translation.

Finally, targeting Mos to modulate pneumonia-induced innate immune memory may represent a promising clinical strategy. On the one hand, blocking pathways that drive HSPC reprogramming could alleviate the harmful effects of trained or tolerized Mos after pneumonia. To date, however, IL-6 remains the only cytokine demonstrated to mediate HSPC reprogramming during a severe viral respiratory infection (42), highlighting the urgent need for further investigations of additional key regulatory signals. On the other hand, although post-pneumonia alterations in AM function could theoretically be modulated by controlling the inflammatory milieu and specific signaling mediators during infection (61,188), the precise mechanisms underlying AM remodeling following pneumonia remain incompletely understood, and the findings are inconsistent. Thus, interventions targeting post-pneumonia AM remodeling remain premature, and future investigations are warranted to address these uncertainties.