Abstract

   Precision management of fibrotic lung diseases is challenging due to
   their diverse clinical trajectories and lack of reliable biomarkers for
   risk stratification and therapeutic monitoring. Here, we validated the
   accuracy of CMKLR1 as an imaging biomarker of the lung
   inflammation-fibrosis axis. By analyzing single-cell RNA sequencing
   datasets, we demonstrated CMKLR1 expression as a transient signature of
   monocyte-derived macrophages (MDMφ) enriched in patients with
   idiopathic pulmonary fibrosis (IPF). Consistently, we identified MDMφ
   as the major driver of the uptake of CMKLR1-targeting peptides in a
   murine model of bleomycin-induced lung fibrosis. Furthermore,
   CMKLR1-targeted positron emission tomography in the murine model
   enabled quantification and spatial mapping of inflamed lung regions
   infiltrated by CMKLR1-expressing macrophages and emerged as a robust
   predictor of subsequent lung fibrosis. Last, high CMKLR1 expression by
   bronchoalveolar lavage cells identified an inflammatory endotype of IPF
   with poor survival. Our investigation supports the potential of CMKLR1
   as an imaging biomarker for endotyping and risk stratification of
   fibrotic lung diseases.
     __________________________________________________________________

   CMKLR1 is an imaging biomarker of the inflammation-fibrosis axis by
   identifying a macrophage subset enriched in lung fibrosis.

INTRODUCTION

   Pulmonary fibrosis is characterized by the replacement of the elastic
   and compliant lung interstitium by a disorganized and collagen-rich
   extracellular matrix (ECM) as a result of various diseases, such as
   infections, immunological diseases, acute respiratory distress syndrome
   (ARDS), and radiation in addition to a diverse category of idiopathic
   interstitial lung diseases (ILDs) ([44]1). This etiological diversity
   along with the variability of the clinical course and treatment
   response of patients has presented major challenges to the precision
   management of fibrotic lung diseases ([45]2). While some patients
   experience nonprogressive disease or gradual physiological
   deterioration over years, others face progressive fibrosis,
   occasionally accompanied by unpredictable episodes of acute
   exacerbations, leading to rapid decompensation and, in some cases,
   death within months ([46]1, [47]3). Despite recent advances in
   antifibrotic and/or immunomodulatory treatments that can mitigate
   disease progression in some patients, current clinical and laboratory
   biomarkers are insufficient for assessing the risk of progressive
   disease, identifying patients who may respond more favorably to various
   therapeutic interventions, or monitoring therapeutic responses
   ([48]2–[49]4). Therefore, diagnostic strategies for noninvasive
   monitoring of pathological drivers of adverse ECM remodeling are vital
   for precision management of fibrotic lung diseases.

   Dysregulated immune responses play a pivotal role in driving adverse
   ECM remodeling across various fibrotic lung diseases, which may be best
   exemplified by fibrosis caused by pneumonia, ARDS, and rheumatological
   diseases ([50]1, [51]4–[52]8). Even in diseases with less robust
   inflammatory profiles, such as idiopathic pulmonary fibrosis (IPF), a
   growing body of evidence supports the role of dysregulated immune
   responses in the pathogenesis of lung fibrosis ([53]4, [54]5, [55]8).
   The accumulation of monocyte-derived macrophages (MDMφ), which replace
   or outnumber tissue-resident alveolar macrophages, has recently emerged
   as a crucial factor in promoting lung fibrosis ([56]4, [57]9–[58]11).
   Under the homeostatic state, tissue-resident alveolar macrophages
   effectively eliminate pathogenic stimuli without provoking an
   inflammatory response ([59]12). However, MDMφ recruited during lung
   injuries continue to exhibit proinflammatory and profibrotic functions
   for extended periods, even after the initial triggering insult has
   resolved, and have emerged as key drivers of adverse ECM remodeling
   ([60]6, [61]7, [62]13, [63]14). Consequently, tracking the in vivo
   kinetics of MDMφ accumulation holds promise to assess the lung
   inflammation-fibrosis axis.

   Positron emission tomography (PET) is a promising technology for
   monitoring dynamic changes in lung immune cell populations. For
   instance, the feasibility of detecting monocyte influx has been shown
   by targeted imaging of C-C chemokine receptor 2 (CCR2) in murine models
   of lung fibrosis and in a first-in-human study of patients with IPF
   ([64]15). However, the long-term retention of MDMφ independent of
   ongoing monocyte influx may not be captured by CCR2-targeting due to
   the rapid down-regulated expression of CCR2 upon monocyte
   differentiation to macrophages ([65]16–[66]18). To address this gap, we
   recently described that targeted imaging of chemokine-like receptor 1
   (CMKLR1), a marker primarily expressed by MDMφ but not by
   tissue-resident alveolar macrophages, enables monitoring of the total
   burden of lung MDMφ in a mouse model of acute lung injury ([67]16). We
   also demonstrated that CMKLR1-expressing lung macrophages
   transcriptionally resemble a recently described subset of CD163^+ and
   LGMN^+ MDMφ with a profibrotic phenotype in patients with COVID-19
   ([68]16, [69]19). In the present study, we investigated the potential
   of CMKLR1-targeted PET to monitor the kinetics of MDMφ accumulation in
   a murine model of fibrotic lung injury and to determine the clinical
   relevance of CMKLR1 as a biomarker for endotyping and prognostication
   of fibrotic lung diseases.

RESULTS

Combined analysis of two single-cell RNA sequencing datasets provides an
updated landscape of macrophages in lung fibrosis

   To investigate the role of CMKLR1 in IPF lung immune cells, we
   performed exploratory data analysis of two publicly available
   single-cell RNA sequencing (scRNA-seq) datasets of whole lungs with
   >300,000 immune cells: Adams et al. ([70]20) includes 28 controls, 32
   IPF, and 18 chronic obstructive pulmonary disease (COPD) lungs;
   Habermann et al. ([71]21) consists of 10 control, 11 IPF, and eight
   assorted ILD lungs ([72]Fig. 1, A to C). This combined sample size
   enabled a more granular perspective of macrophage subpopulations in
   human lung fibrosis than previously reported.

Fig. 1. scRNA-seq analysis of human lungs in disease and health.

   [73]Fig. 1.
   [74]Open in a new tab

   (A) Waffle plots representing the number of subjects per disease state,
   per dataset (top) and UMAP of cells labeled by subject (bottom). (B)
   UMAP of cells labeled by cell type. (C) UMAP and (D) heatmap of
   normalized gene expression across cell types, datasets, and diseases.
   Each column represents the average gene expression for one subject per
   cell type. The upper section corresponds to sensitive and specific
   markers per cell type; the lower section corresponds to either broadly
   observed or disease-restricted features. Values are min-max normalized
   within each dataset. (E) Partition-based graph abstraction (PAGA) of
   subclustered myeloid cells in a layout corresponding to the earlier
   UMAP with vertices colored by their respective cell type. All edges
   have a confidence value >0.65. (F) PAGA with average normalized
   expression of CMKLR1. (G) PAGA with relative disease portion makeup.
   (H) IPF and control representation across two PAGA subgraph
   trajectories ordered from cMonocyte toward IPF-enriched macrophage
   archetype. The Mann-Whitney U test was used to compare the proportion
   of myeloid cells in each cluster between IPF and control lungs.
   Independent tests were applied to each dataset; *P < 0.05, **P <0.01,
   ***P < 0.001. (I to J) Generalized additive model (GAM) fits of gene
   expression as a function of diffusion pseudotime for each
   aforementioned trajectory. Line colors correspond to cMonocyte features
   (yellow), FOLR2 macrophage features (cyan), CMKLR1 (red), and IPF
   features (magenta) with corresponding colors in above heatmap.

   Macrophage phenotypes were first broadly classified into two main
   groups: alveolar and nonalveolar. Alveolar macrophages were defined by
   expression patterns of genes, including FABP4, LSAMP, and PPARG
   ([75]Fig. 1D and data file S1), consistent with previous descriptions
   ([76]22). We further subdivided these cells into Alv-Mφ (FABP4^Hi) and
   FABP4^Int-Mφ, where the latter had relatively lower expression of the
   same distinguishing features (fig. S1). Alv-Mφ and FABP4^Int-Mφ were
   transcriptomically distinguishable phenotypes that co-occurred in every
   sample, irrespective of dataset or disease state ([77]Fig. 1D) with
   sufficient differences in global gene expression patterns to separate
   these populations in the feature-space of the uniform manifold
   approximation and projection (UMAP) layout. However, the dynamic
   relationship between FABP4^Int-Mφ and FABP4^Hi-Alv-Mφ, such as a
   transitioning state between the two subsets or a stable subset observed
   in a transient state of activity, remains unknown.

   The remaining group of nonalveolar macrophages was further divided into
   three groups. The largest and most indeterminate group was SPP1-Mφ,
   which had elevated SPP1, PLA2G7, and SDC2, but were otherwise
   classified by exclusion. IPF profibrotic macrophages were a pronounced
   archetype among SPP1-Mφ ([78]20, [79]22) ([80]Fig. 1, A and D); a
   separate inflammatory archetype from a single control outlier also
   belonged to this SPP1-Mφ ([81]20) ([82]Fig. 1A).

   Distinguishing features of FOLR2-Mφ include LYVE1, F13A1, SEPP1,
   LILRB5, and CCL13. Recent studies have brought attention to the
   ubiquity of FOLR2-Mφ throughout the body ([83]23) and the role they
   play in the lung following injury ([84]24). The role of FOLR2-Mφ has
   been overlooked in previous scRNA-seq studies of interstitial lung
   disease ([85]20–[86]22) and their status in IPF remains poorly
   characterized.

   We also identified a rare, third population of nonalveolar
   macrophage–labeled OLFML3-Mφ. These cells were classified as
   macrophages based on the shared expression of MERTK, PLA2G7, and C1Q
   complex genes, though they also have dendritic features like FCGR2B,
   FCER1A, and elevated MHC-II ([87]Fig. 1D and data file S1). In the
   context of other macrophages, these cells share the expression of
   SLC40A1 and OLFM2B with FOLR2-Mφ but otherwise lack the expression of
   the remaining FOLR2-Mφ expression suite. Across samples in both
   datasets, OLMFL3-Mφ cells uniquely express C3, FCGBP, PLD4, GPR34,
   IGSF21, and OLFML3 ([88]Fig. 1D and data file S1). These cells were
   commonly found across IPF samples and only rarely found in controls
   ([89]Fig. 1, A and D).

CMKLR1 expression is a hallmark of nascent MDMφ enriched in pulmonary
fibrosis and gradually declines as macrophages acclimate to the fibrotic
microenvironment

   Between both datasets, CMKLR1 expression was most frequently detected
   across macrophage populations ([90]Fig. 1, C and D, and data file S2),
   consistent with our previous finding in patients with COVID-19
   ([91]16). Irrespective of the dataset or whether the cells were from
   IPF or control lungs, FOLR2-Mφ had the highest detection frequency and
   average normalized expression of CMKLR1 across all immune cell types.
   The next highest ranked cell types in descending order were OLFML3-Mφ,
   SPP1-Mφ, FABP4^Int-Mφ, and, lastly, FABP4^Hi-Alv-Mφ (data file S2).
   These findings are consistent with our previous analyses where
   nonalveolar macrophages expressing CMKLR1 in COVID-19 lungs also
   expressed FOLR2-Mφ features like LYVE1 and F13A1 ([92]16) ([93]Fig. 1D)
   and how CMKLR1 expression was observed highest among FOLR2^+ and SPP1^+
   interstitial macrophage in patients with cystic fibrosis ([94]16).

   An exploratory analysis of CMKLR1 expression in different diseases
   suggested a preferential expansion of CMKLR1-expressing macrophages in
   IPF compared to other ILD entities (i.e., chronic hypersensitivity
   pneumonitis, nonspecific interstitial pneumonia, unclassifiable ILD,
   and sarcoidosis) and COPD (fig. S2). However, considering the
   relatively lower representation of lung cells from these disease states
   compared to IPF, further studies are required to characterize
   disease-specific patterns of CMKLR1 expression in various ILDs and in
   different endotypes or stages of COPD.

   Changes in CMKLR1 expression were not commensurate with the boundaries
   of nonalveolar macrophage cell types: we observed more CMKLR1 variation
   among SPP1-Mφ than between SPP1-Mφ and adjacent macrophages ([95]Fig.
   1, B and C). Summarizing expression values across cell types in this
   context has limitations. Qualitatively, we observed CMKLR1 expression
   peaking across a variety of nonalveolar macrophages near their
   boundaries with the monocyte/cDC2/macrophage intermediate population
   “junction.” Expression gradually declined in a diffuse pattern where
   ultimately the most extremal cell types and cell states have low-to-no
   detectable CMKLR1 ([96]Fig. 1C). This pattern indicates that CMKLR1
   expression is broadly associated with nascent MDMφ.

   Considering that IPF severity is typically worse in the lung bases, we
   compared CMKLR1 expression by immune cells between the upper and lowers
   lobes of patients with IPF versus control participants using the
   scRNA-seq dataset from Morse et al. ([97]25) [Gene Expression Omnibus
   (GEO) accession [98]GSE128033]. Consistent with our results from the
   Adams et al. ([99]20) and Habermann et al. ([100]21) datasets, CMKLR1
   expression was highest in SPP1-Mφ and FOLR2-Mφ with a low expression in
   Alv-Mφ and extremal profibrotic SPP1-Mφ (i.e., SPP1^+ macrophages with
   the highest intensity of the profibrotic gene expression) (fig. S3).
   Upper lobes from patients with IPF had more FABP4^+ alveolar
   macrophages and fewer IPF-associated profibrotic SPP1^+ macrophages
   than the lower lobes but similarly elevated numbers of CMKLR1^+
   macrophages (IPF upper lobe: 8.4% and lower lobe: 8.9% versus control
   upper lobe: 2.3% and lower lobe: 0.8%). While this requires further
   investigation, it may be speculated that accumulation of CMKLR1^+
   macrophages in the lungs may represent a more reliable biomarker of
   early disease than the SPP1^+ profibrotic macrophages described in
   end-stage disease.

   To prepare for trajectory-based analysis, we assessed the robustness of
   cell-cell relationships to ensure that downstream pseudotime orderings
   would be constrained to plausible outcomes. Monocytes, macrophages, and
   dendritic cells contiguous in UMAP space ([101]Fig. 1B) were isolated
   and reclustered with a high-resolution parameter yielding over 50
   granular subclusters organized below the level of cell types. Cell-cell
   connections between these subpopulations were statistically modeled
   with partition-based graph abstraction (PAGA) ([102]26) and
   intercluster connections with high confidence values were preserved for
   downstream analysis ([103]Fig. 1E).

   Plotting a trajectory from monocytes toward extremal macrophage fates,
   we observed a branching event between FOLR2-Mφ and SPP1-Mφ that
   occurred among junction clusters ([104]Fig. 1E), consistent with how
   differentiation occurs in adult mouse lungs ([105]24), in which Folr2^+
   interstitial macrophages are gradually replaced by circulating
   monocytes over time. Although under healthy conditions Folr2^+
   interstitial macrophages have a longer half-life than other
   interstitial macrophages in adult mice ([106]24, [107]27, [108]28),
   their rate of turnover under fibrotic conditions is unknown.
   Unexpectedly, one FOLR2-Mφ subcluster featuring the highest overall
   CMKLR1 expression ([109]Fig. 1F) was overrepresented in IPF samples
   ([110]Fig. 1G) and had several high confidence connections directly to
   subclusters of the SPP1-Mφ IPF profibrotic archetype ([111]Fig. 1E).
   These patterns suggest that after recruited lung monocytes diverge
   toward either SPP1-Mφ or FOLR2-Mφ states, IPF macrophages from either
   population may converge into a common profibrotic SPP1-Mφ archetype,
   although this should be interpreted within the constraints of
   pseudotime analysis, which cannot distinguish bona fide phenotypic
   transitions from phenotypic similarities.

   We then focused on two putative trajectories that both start with
   classical monocytes (cMonocytes) and end with profibrotic SPP1-Mφ, one
   via FOLR2-Mφ and the other via SPP1-Mφ ([112]Fig. 1H). Following the
   FOLR2-Mφ PAGA trajectory, normalized CMKLR1 expression peaked in a
   subcluster of FOLR2-Mφ, which was overrepresented across IPF lungs in
   both datasets ([113]Fig. 1H). Generalized additive models (GAMs) that
   fit the corresponding cell-level pseudotime trajectory showed that
   CMKLR1 expression is rising concurrently with the decline of monocyte
   features, peaking simultaneously with FOLR2 as profibrotic gene
   expression initializes, then gradually declining as FOLR2 is lost and
   profibrotic features gradually rise ([114]Fig. 1I). Similarly, the
   trajectory restricted to SPP1-Mφ had CMKLR1 expression peaking while
   monocyte features start declining and expression of profibrotic
   features begin to rise ([115]Fig. 1J). CMKLR1 was represented in these
   models as a transient hallmark of MDMφ among both FOLR2-Mφ and SPP1-Mφ
   states, and that CMKLR1 expression gradually declines as macrophages
   acclimated to the fibrotic microenvironment.

MDMφ are the major driver of the uptake of CMKLR1-targeting peptides in a
murine model of fibrotic lung injury

   We recently reported the feasibility of targeted imaging of CMKLR1 in
   acute lung injury using a proteolytically stable and high-affinity
   copper-64–labeled radiotracer ([^64Cu]NODAGA-CG34) ([116]16) derived
   from an analog of the C-terminus of chemerin ([117]29). To leverage the
   advantage of flow cytometry for quantifying the abundance of various
   CMKLR1-expressing immune cells and their respective contribution to
   radiotracer uptake during different stages of fibrotic lung injury, we
   synthesized fluorochrome-labeled analogs of [^64Cu]NODAGA-CG34 by
   replacing the chelator moiety (NODAGA) of the tracer with fluorescein
   (6CF-CG34) or TAMRA (6TAM-CG34) (fig. S4). We confirmed that the
   fluorochrome-modified CG34 peptides exhibited similar pharmacodynamic
   properties as NODAGA-CG34, including comparable binding/uptake by mouse
   peritoneal macrophages, which constitutively express CMKLR1 ([118]16)
   and Chinese hamster ovary (CHO) cells transiently transfected with
   murine CMKLR1 (mCMKLR1), as well as displacement by NODAGA-CG34 in
   competitive uptake assays in peritoneal macrophages (fig. S5, A to H).
   Furthermore, we established that CG34-derived peptides are internalized
   in a CMKLR1-selective manner (fig. S5I) with no significant
   binding/uptake by the two related G protein–coupled receptors of
   chemerin, namely, chemokine (C-C motif) receptor–like 2 (CCRL2) and G
   protein–coupled receptor 1 (GPR1) (fig. S5, F and G). We also
   demonstrated the high affinity of fluorochrome-conjugated CG34 to human
   CMKLR1 (fig. S5H), using transiently transfected CHO cells, which was
   comparable to its affinity to mouse CMKLR1.

   Using the gating strategy outlined in fig. S6, we identified and
   quantified major lung immune cell populations in control mice and at
   different time points (1, 2, and 4 weeks) after intratracheal bleomycin
   administration (fig. S7). These time points approximately correspond to
   the peak pneumonitis phase, early ECM remodeling with continued
   pneumonitis, and the resolution of inflammation with established
   fibrosis, respectively ([119]30). We were particularly interested in
   characterizing different macrophage subsets as our recent work revealed
   that increased uptake of [^64Cu]NODAGA-CG34 in a preclinical model of
   acute lung injury was primarily driven by MDMφ ([120]16). To this end,
   we used a previously established flow cytometric method to distinguish
   MDMφ (CD11b^hi/CD11c^hi/SiglecF^int) from tissue-resident alveolar
   macrophages (CD11b^lo/CD11c^hi/SiglecF^hi), as validated by lineage
   tracing studies ([121]6). At baseline, tissue-resident alveolar
   macrophages were the predominant population recovered from the lungs
   with fewer interstitial macrophages and negligible MDMφ ([122]Fig. 2, A
   and B). It is noteworthy that the low abundance of interstitial
   macrophages at baseline is a reflection of the lower recovery of these
   cells compared to alveolar macrophages during single-cell preparation
   and the number of interstitial macrophages is much higher in lung
   parenchyma as demonstrated by stereological studies compared to flow
   cytometry ([123]31). At 1 week following bleomycin, a significant
   accumulation of MDMφ was observed, whereas the number of alveolar
   macrophages remained stable ([124]Fig. 2, A and B). By the second week,
   the number of MDMφ remained similar to that at 1 week, but there was a
   substantial increase in the number of alveolar macrophages. MDMφ at 2
   weeks postbleomycin showed increasing resemblance to tissue-resident
   alveolar macrophages (i.e., increased SiglecF and decreased CD11b
   expression) ([125]Fig. 2A), indicating their progressive
   differentiation toward the alveolar macrophage phenotype. By the fourth
   week, the number of both MDMφ and alveolar macrophages returned to
   baseline levels. In contrast, the number of lung monocytes (Ly6C^hi and
   Ly6C^lo) remained relatively stable from week 1 to week 4 after
   bleomycin-induced injury.

Fig. 2. Flow cytometric identification of cells contributing to the uptake of
CMKLR1-targeting peptide.

   [126]Fig. 2.
   [127]Open in a new tab

   (A) Alveolar macrophages, interstitial macrophages, and MDMφ were
   identified based on differential expressions of CD11b, CD11c, and
   SiglecF. (B) Kinetic changes in the absolute number of recovered
   monocyte and macrophage subsets revealed a marked infiltration of the
   lungs during the first 2 weeks after bleomycin administration,
   partially resolved by week 4. (C) Representative histograms reveal the
   uptake of CMKLR1-targeting fluorescent ligand 6CF-CG34 by various lung
   monocyte/macrophage populations in bleomycin- or PBS-treated mice
   [gray: background/autofluorescence; red: total binding of 6CF-CG34 in
   the absence of Chem[145–157] (a high-affinity CMKLR1 ligand derived
   from the amino acids 145 to 157 of chemerin used as a competitive
   blocker of 6CF-CG34 uptake); blue: nonspecific binding of 6CF-CG34 in
   the presence of 10 μM Chem[145–157]]. (D) Quantification of specific
   (total minus nonspecific) uptake of 6CF-CG34 by selected monocyte and
   macrophage populations. MDMφ demonstrated the highest 6CF-CG34 uptake
   at 1 and 2 weeks postbleomycin, while interstitial macrophages
   demonstrated a transient increase in 6CF-CG34 uptake at 1 week
   postbleomycin. A stacked-bar chart summary of cell counts for CMKLR1
   expressing populations multiplied by the cellular uptake of 6CF-CG34
   (molecules per cell) (E) showed that MDMφ accounted for most of the
   lung uptake of 6CF-CG34 at 1 and 2 weeks following bleomycin. n = 3
   male and 3 female mice for control and 1- and 2-week postbleomycin
   groups; n = 4 male and 4 female mice for the 4-week group. aMφ,
   alveolar macrophages; iMφ, interstitial macrophages; NK cells, natural
   killer cells. Comparisons were made between various treatment time
   points to the control group for alveolar macrophages, interstitial
   macrophages, Ly6C^hi monocytes, and Ly6C^lo monocytes or to 1-week
   treatment group for MDMφ (as the control group is omitted). *P < 0.05,
   **P < 0.01, ***P < 0.001, and ****P < 0.0001. Statistical significance
   was calculated using a one-way ANOVA with a post-hoc two-tailed
   Fisher’s exact test.

   We observed a high uptake of 6CF-CG34 by MDMφ during the initial 2
   weeks following bleomycin-induced injury, which trended toward a
   decrease by the fourth week ([128]Fig. 2, C and D). Interstitial
   macrophages displayed only a transient increase in 6CF-CG34 uptake at 1
   week, which returned to the baseline level at 2 and 4 weeks
   postbleomycin. However, alveolar macrophages and monocytes (Ly6C^hi and
   Ly6C^lo) did not exhibit an appreciable specific uptake of 6CF-CG34
   throughout the 4-week period following injury. The only other immune
   cell type that exhibited significant 6CF-CG34 uptake was natural killer
   cells, but their abundance or CMKLR1 expression remained unchanged
   after bleomycin-induced injury (fig. S7).

   To estimate the relative contribution of different immune cell
   populations to the total uptake of the fluorescent CMKLR1-selective
   ligand, we multiplied the cell count of the three major
   CMKLR1-expressing cell types (MDMφ, interstitial macrophages, and
   natural killer cells) by the number of 6CF-CG34 molecules taken up per
   cell ([129]Fig. 2E). At baseline, we observed low 6CF-CG34 uptake.
   However, at 1 and 2 weeks postinjury, there was a significant increase
   in total 6CF-CG34 uptake, primarily driven by the accumulation of MDMφ,
   which accounted for 78 and 86% of the observed uptake by
   monocyte/macrophages and natural killer cells at 1 and 2 weeks,
   respectively. By 4 weeks, the 6CF-CG34 uptake had decreased to
   approximately baseline levels, reflecting a reduction in the abundance
   of MDMφ and their CMKLR1 expression level. Natural killer cells made a
   minimal contribution to the total 6CF-CG34 uptake in bleomycin-injured
   lungs, constituting only 11% at 1 and 2 weeks compared to 95% at
   baseline, due to their markedly lower abundance compared to MDMφ in
   injured lungs.

[^64Cu]NODAGA-CG34 PET monitors lung inflammation throughout the progression
of bleomycin-induced fibrotic injury

   Building on our flow cytometry findings, we investigated the potential
   of CMKLR1-targeted PET using [^64Cu]NODAGA-CG34 to noninvasively track
   the kinetics of MDMφ accumulation in bleomycin-induced lung injury.
   Kinetic PET studies in 1 week postbleomycin versus control mice
   revealed substantial blood pool radioactivity during the initial 30 min
   after tracer administration hindering visualization and accurate
   quantification of [^64Cu]NODAGA-CG34 uptake in the lungs ([130]Fig. 3,
   A and B, and fig. S8). There was a progressive clearance of blood pool
   activity from 30 to 60 min, reaching a near plateau from 60 to 90 min.
   Therefore, for the remainder of the study, radiotracer quantifications
   were performed on static scans ~90 min post-tracer.

Fig. 3. [^64Cu]NODAGA-CG34 PET/CT in a murine model of bleomycin-induced lung
fibrosis.

   [131]Fig. 3.
   [132]Open in a new tab

   (A and B) Kinetic [^64Cu]NODAGA-CG34 uptake in the lungs and blood pool
   activity, quantified as the maximal % injected dose (ID)/ml[max] (A)
   and mean %ID/ml[mean] (B), over the first 90 min following radiotracer
   injection to bleomycin-treated (week 1, n = 3) and control (n = 2)
   mice. At 90 min, there was a high accumulation of [^64Cu]NODAGA-CG34 in
   the lungs of bleomycin-treated versus control mice, while the
   radiotracer was mostly cleared from the circulation. (C) The design of
   the time-course study for quantifying [^64Cu]NODAGA-CG34 uptake at
   distinct stages of bleomycin-induced fibrotic lung injury. (D)
   Representative axial CT, PET, and coregistered PET/CT images acquired
   ~90 min following administration of [^64Cu]NODAGA-CG34 to a control
   mouse or mice at 1, 2, or 4 weeks postbleomycin injury (scale bar, 5
   mm). Diffuse lung uptake of [^64Cu]NODAGA-CG34 was observed in the
   lungs of mice at 1 and 2 weeks postbleomycin, which returned to near
   baseline by 4 weeks. The specificity of radiotracer uptake was
   confirmed by blockade of [^64Cu]NODAGA-CG34 uptake in the lungs when
   co-injected with ~100-fold molar excess of nonradiolabeled NODAGA-CG34.
   PET-derived quantification demonstrated significant increases in the
   lung uptake of [^64Cu]NODAGA-CG34 both measured as %ID/ml[max] (E) and
   %ID/ml[mean] (F) at 1 and 2 weeks postbleomycin. %ID/ml[max] and
   %ID/ml[mean] data in (A), (B), (E), and (F) represent the average
   values of the left and right lungs for each mouse. Ctl, control; Bleo,
   bleomycin. Statistical significance between groups was calculated using
   a one-way ANOVA with a two-tailed Fisher’s exact post hoc test.

   In line with flow cytometric results, visual examination of PET
   demonstrated increased lung uptake of [^64Cu]NODAGA-CG34 at 1 and 2
   weeks following bleomycin, which decreased to levels nearly similar to
   those of control mice by 4 weeks ([133]Fig. 3, C and D). Quantitative
   analysis of PET corroborated our qualitative assessment by
   demonstrating significant increases in the lung uptake of
   [^64Cu]NODAGA-CG34 at 1 and 2 weeks postbleomycin both in terms of the
   uptake in the maximally affected lung regions [i.e., maximal percent
   injected dose per ml (%ID/ml[max]); [134]Fig. 3E] and the average
   global uptake throughout the lungs [i.e., mean percent injected dose
   per ml (%ID/ml[mean]); [135]Fig. 3F]. We confirmed the specificity of
   radiotracer uptake by demonstrating the blockade of the lung uptake of
   [^64Cu]NODAGA-CG34 upon co-injection of excess nonradiolabeled
   NODAGA-CG34 ([136]Fig. 3, D and F).

   Quantification of radiotracer uptake using ex vivo γ-counting, as the
   gold standard method, yielded similar results with significant
   increases in radiotracer uptake at 1 and 2 weeks postbleomycin
   ([137]Fig. 4A). There was also a strong correlation between in vivo
   PET-derived and ex vivo γ-counting–derived measurements, which
   establishes the accuracy of PET-derived quantification of
   [^64Cu]NODAGA-CG34 uptake in the lungs (R^2 = 0.77 and P < 0.0001)
   ([138]Fig. 4B). In addition to the lungs, significant specific uptake
   of [^64Cu]NODAGA-CG34 was present in several CMKLR1-rich organs,
   including liver, spleen, and thymus ([139]Fig. 4, C and D).

Fig. 4. Ex vivo measurement of [^64Cu]NODAGA-CG34 uptake and biodistribution.

   [140]Fig. 4.
   [141]Open in a new tab

   (A) Ex vivo γ-counting confirmed significantly increased lung uptake of
   [^64Cu]NODAGA-CG34 at 1 and 2 weeks postbleomycin, which returned to
   baseline values at week 4. (B) The strong correlation between in vivo
   PET-derived and ex vivo γ-counting–derived measurements of
   [^64Cu]NODAGA-CG34 uptake in the lungs confirms the accuracy of PET for
   noninvasive quantification of tracer uptake. (C) Biodistribution of
   [^64Cu]NODAGA-CG34 was determined by γ-counting–derived of harvested
   organs (* or # P < 0.05; ** or ## P < 0.01; *** or ### P < 0.001; *
   represents comparison between control and nonblocked groups; #
   represents comparison between 2-week and blocked groups). (D)
   Representative full body PET/CT maximal intensity projections (scale
   bars, 9 mm) and magnified views of the thorax (yellow boxes, scale
   bars, 4 mm) to visualize [^64Cu]NODAGA-CG34 uptake in the lung and
   other organs. Statistical significance between groups in (A) and (C)
   was calculated using a one-way ANOVA with a two-tailed Fisher’s exact
   post hoc test.

[^64Cu]NODAGA-CG34 PET delineates inflamed lung regions infiltrated by
CMKLR1-expressing macrophages

   To confirm the accuracy of [^64Cu]NODAGA-CG34 PET to provide a spatial
   map of inflamed lung tissues infiltrated by CMKLR1-expressing
   macrophages, we performed autoradiography of the lungs after completion
   of PET and conducted subsequent immunofluorescence staining for CMKLR1
   and the macrophage marker F4/80 ([142]Fig. 5). There was a robust
   colocalization between tracer uptake as detected by in vivo PET and the
   corresponding ex vivo autoradiography ([143]Fig. 5, A and B),
   confirming the accuracy of noninvasive spatial localization of tracer
   uptake. The regions exhibiting high [^64Cu]NODAGA-CG34 uptake
   corresponded to inflamed areas with an abundance of CMKLR1-expressing
   F4/80^+ macrophages ([144]Fig. 5, C and D). These results collectively
   indicate that [^64Cu]NODAGA-CG34 PET enables quantitative monitoring
   and spatial mapping of MDMφ recruitment at various stages of
   bleomycin-induced fibrotic lung injury.

Fig. 5. Colocalization of [^64Cu]NODAGA-CG34 uptake with inflamed lung
regions infiltrated by CMKLR1-expressing macrophages.

   [145]Fig. 5.
   [146]Open in a new tab

   Representative PET/CT images (A) from a bleomycin-treated (1 week)
   mouse reformatted to the plane of ex vivo autoradiography (B) and
   hematoxylin and eosin (H&E)–stained (C) and immunostained tissues (D)
   demonstrate colocalization of [^64Cu]NODAGA-CG34 uptake with inflamed
   lung regions densely infiltrated by CMKLR1-expressing F4/80^+
   macrophages (pink box). Regions of low [^64Cu]NODAGA-CG34 uptake
   (yellow box) demonstrate minimal expression of CMKLR1. Scale bars in
   CT, PET/CT, and autoradiography images: 4 mm; in H&E staining: 5-mm; in
   5× immunofluorescence images: 800 μm; and in 20× immunofluorescence
   images: 200 μm. n = 3 male mice.

[^64Cu]NODAGA-CG34 PET predicts progressive fibrotic lung injury in the
murine model

   To determine the potential of CMKLR1-targeted PET to predict the
   progression of pulmonary fibrosis, we conducted a longitudinal study in
   which mice underwent PET/CT at week 1 postbleomycin and were followed
   until week 4 to assess the extent of fibrosis ([147]Fig. 6A). As shown
   by representative images in [148]Fig. 6B, regions of lung with high
   [^64Cu]NODAGA-CG34 uptake at week 1 postbleomycin corresponded to
   regions that subsequently developed severe fibrosis as detected by
   high-resolution computed tomography (CT) conducted at week 4. Moreover,
   there was a significant correlation between PET-derived
   [^64Cu]NODAGA-CG34 uptake at week 1 and the extent of lung fibrosis at
   week 4, as determined by their hydroxyproline content ([149]Fig. 6C)
   and increased lung attenuation ([150]Fig. 6D). In contrast, the
   severity of airspace opacification as measured by increased lung
   attenuation at week 1 postbleomycin only trended toward a weak
   correlation with the hydroxyproline content at week 4 (R^2 = 0.17 and P
   = 0.087; [151]Fig. 6E) and did not significantly correlate with the
   increased lung attenuation ([152]Fig. 6F). Together, these results
   confirm the accuracy of [^64Cu]NODAGA-CG34 uptake during the early
   stage of lung injury to predict the severity and distribution of future
   fibrosis, which complements the anatomical information obtained by CT.

Fig. 6. Prognostication of fibrotic lung injury by [^64Cu]NODAGA-CG34 PET
during the early stage of bleomycin-induced injury.

   [153]Fig. 6.
   [154]Open in a new tab

   (A) Experimental design to determine the potential of
   [^64Cu]NODAGA-CG34 PET during the inflammatory phase of
   bleomycin-induced injury in predicting the severity of future lung
   fibrosis. (B) Representative PET/CT images (scale bar, 5 mm)
   demonstrate minimal lung uptake in a control mouse at 1 week following
   intratracheal PBS administration (top row) and no CT evidence of
   fibrosis at week 4. Examples from bleomycin-treated mice with low
   (middle row) versus high (lower row) tracer uptake demonstrate that
   [^64Cu]NODAGA-CG34 uptake at week 1 colocalizes and subjectively
   correlates with the severity of future fibrosis at week 4 as visualized
   by high-resolution CT (HRCT). The severity of pulmonary fibrosis at 4
   weeks, measured by the hydroxyproline contents (C) or increased lung
   attenuation (D), correlated with [^64Cu]NODAGA-CG34 uptake in the
   corresponding lungs as measured by %ID/ml[mean] at 1 week (Pearson
   correlation along with 95% confidence intervals). However, the
   increased lung attenuation at week 1 postbleomycin only trended toward
   a weak correlation with the hydroxyproline content at week 4 (E) and
   did not significantly correlate with the increased lung attenuation
   (F).

High CMKLR1 expression by BAL cells identifies an inflammatory IPF endotype
with a poor prognosis

   To explore the potential of CMKLR1 as a prognostic and endotyping
   biomarker, we conducted a secondary analysis of harmonized data (fig.
   S9) from a longitudinal cohort of IPF patients at three sites:
   Freiburg, Leuven, and Siena ([155]32). We first compared CMKLR1
   expression in bronchoalveolar lavage (BAL) cells in healthy controls
   versus patients with IPF categorized based on their
   gender-age-physiology (GAP) index, a widely accepted clinical risk
   stratification model ([156]33). We observed a higher CMKLR1 expression
   in patients with IPF compared to controls, but no significant
   difference was present among patients in different GAP score categories
   ([157]Fig. 7A). Notably, there was a marked heterogeneity in CMKLR1
   expression in patients with IPF independent of their GAP index.

Fig. 7. CMKLR1 expression by BAL cells identifies an inflammatory and poor
prognostic endotype of IPF.

   [158]Fig. 7.
   [159]Open in a new tab

   (A) CMKLR1 expression by BAL cells is increased in patients with IPF
   compared to healthy controls but is not different across different
   categories of GAP scores. (B) Volcano plot of differentially expressed
   genes demonstrates numerous differentially expressed genes, including
   multiple markers of inflammation and ECM remodeling, in patients with
   high versus low CMKLR1 expression. (C) Pathway enrichment analysis
   confirms significant enrichment of inflammatory and ECM remodeling
   pathways in patients with high versus low CMKLR1 expression. (D)
   Kaplan-Meier survival curves based on CMKLR1 expression alone (left) or
   after stratification by GAP score (E and F) demonstrates that high
   CMKLR1 expression predicts poor survival. (G and H) ROC curves
   comparing CMKLR1 expression versus GAP score for predicting patient
   mortality after indicated time points following BAL.

   To investigate whether the heterogeneity of CMKLR1 expression may be
   leveraged to identify distinct inflammatory endotypes of IPF, we
   analyzed the BAL transcriptome of patients with CMKLR1 expression above
   versus below the median value, which revealed numerous differentially
   expressed genes between the two groups ([160]Fig. 7B and data file S3).
   Notably, patients with high CMKLR1 expression had a significantly
   higher expression of many genes with established roles in inflammation,
   including chemokines/chemokine receptors (e.g., CCL2, CCL7, CXCL14, and
   CXCR4) and cytokines/cytokine receptors (e.g., IL1R2 and IL12RB2), as
   well as ECM remodeling (e.g., LGMN, MMP25, SPP1, COL1A1, and FGFBP2)
   ([161]Fig. 7B). Pathway enrichment analysis of the differentially
   overexpressed genes confirmed the involvement of inflammatory and ECM
   remodeling pathways in patients with high CMKLR1 expression ([162]Fig.
   7C).

   Next, we assessed the potential of CMKLR1 expression as a prognostic
   biomarker by comparing the survival of patients in different tertiles
   of CMKLR1 expression. Notably, higher CMKLR1 expression was associated
   with a progressively worse prognosis with median survival of 388 days
   in the highest tertile versus 718 days in the middle and 1394 days in
   the lowest tertiles ([163]Fig. 7D). Cox proportional hazards analysis
   adjusted for potential confounders confirmed that CMKLR1 expression
   predicts mortality independent of age, sex, study site, and GAP score
   (fig. S10). We also explored the prognostic value of two additional
   biomarkers along the monocyte-to-macrophage differentiation spectrum,
   CCR2, as a monocytic marker and SPP1, as a hallmark of differentiated
   profibrotic macrophages. Intriguingly, high expression levels of both
   biomarkers were significantly associated with a worse survival (fig.
   S11). The divergence in survival curves for SPP1 across the tertiles
   was overall similar to that of CMKLR1 with a median survival of 432
   days in the highest tertile compared to 958 days in the middle and 1404
   days in the lowest tertiles. In contrast, the separation of the
   survival curves was less distinct for CCR2 with the median survival of
   586 days in the highest tertile versus 840 days in the middle and 1272
   days in the lowest tertiles. These findings support that different
   biomarkers within the monocyte-to-macrophage differentiation spectrum
   offer valuable prognostic insights in IPF.

   Furthermore, we confirmed the potential of CMKLR1 expression to provide
   incremental risk stratification information beyond the existing
   clinical tools by demonstrating that CMKLR1 expression was associated
   with a worse survival in analyses performed after stratification of
   patients to low GAP index (0 to 4; [164]Fig. 7E) and high GAP index (5
   to 8; [165]Fig. 7F). We also performed receiver operating
   characteristic (ROC) analysis to compare the performance of CMKLR1
   expression and GAP index in predicting mortality risk at different time
   points after BAL ([166]Fig. 7, G and H). The ROC analysis revealed that
   the CMKLR1 expression had a higher area under the curve than the GAP
   score during the first 2 years, indicating its superior predictive
   ability over short-to-intermediate follow-up periods, whereas the GAP
   score was a better predictor of long-term (i.e., years 3 and 4)
   survival.

   Together, these findings support that the heterogeneous CMKLR1
   expression may be harnessed to identify a distinct endotype of IPF with
   a more intense inflammatory and ECM remodeling profile associated with
   a high risk of progression and mortality over a short-to-intermediate
   interval.

CMKLR1 is expressed in different categories of human fibrotic lung diseases

   Motivated by our finding of CMKLR1 mRNA expression by lung macrophages
   and BAL cells in fibrotic lung diseases, we performed an exploratory
   analysis to confirm histological evidence of CMKLR1 protein expression
   in the lungs of patients with IPF and other fibrotic lung diseases
   using an existing biorepository (table S1). We observed the presence of
   CMKLR1-expressing leukocytes in the lungs of patients with IPF,
   sarcoidosis, scleroderma, rheumatoid arthritis, and coal worker
   pneumoconiosis (n = 3 per disease category) ([167]Fig. 8 and fig. S12).
   It is noteworthy that CMKLR1 expression was also detectable in
   CD45-negative cells, consistent with the scRNA-seq data, indicating the
   presence of CMKLR1 expression in nonimmune cells, albeit at a lower
   abundance when compared to immune cells (fig. S13). Together, our
   exploratory histology evaluation suggests that CMKLR1 is expressed in
   various fibrotic lung diseases. However, it should be noted that
   explanted lungs usually represent features of end-stage disease, which
   may not be reflective of the initial pathology. Therefore, additional
   studies are necessary to elucidate the disease-specific patterns of
   CMKLR1 expression in these conditions and to ascertain its potential
   utility as a clinical biomarker for disease endotyping and risk
   stratification.

Fig. 8. CMKLR1 expression in the lungs of patients with fibrotic lung
diseases.

   [168]Fig. 8.
   [169]Open in a new tab

   Low-magnification bright-field and high-magnification immunostaining
   for CMKLR1 and CD45 demonstrate the abundance of CMKLR1-expressing
   leukocytes in the lungs of patients with pulmonary fibrosis associated
   with various etiologies. While there was a marked heterogeneity in the
   expression of CMKLR1 in different regions of the lungs, the example
   images were selected to convey CMKLR1 expression in diseased regions of
   the lungs. CMKLR1-expressing CD45-negative nonimmune cells were also
   noted across different disease categories. n = 3 patients for each
   disease category (examples from the other two patients in each category
   of lung diseases are shown in fig. S13). CW, coal work pneumoconiosis;
   RA-ILD, rheumatoid arthritis–associated interstitial lung disease.
   Scale bars are 500 μm in bright-field images and 200 μm in fluorescent
   images.

DISCUSSION

   In this study, we established CMKLR1 as a biomarker of MDMφ enriched in
   patients with fibrotic lung diseases, and demonstrated the accuracy of
   CMKLR1-targeted PET to noninvasively monitor the accumulation of MDMφ
   and predict the severity of future fibrosis in the murine model of
   bleomycin-induced lung injury. Furthermore, we established the
   translational relevance of our findings by demonstrating that high
   expression of CMKLR1 in BAL cells identifies an inflammatory endotype
   of IPF characterized by a poor prognosis and a transcriptional profile
   indicative of heightened inflammatory and ECM remodeling pathways.

   Current therapeutic approaches to fibrotic lung diseases do not take
   into account the molecular endotypes and heterogeneity of the
   underlying pathogenic processes, which drive adverse ECM remodeling. As
   a result, these approaches heavily rely on long-term monitoring of
   functional and structural consequences, for example, by pulmonary
   function tests and CT, to assess disease progression and treatment
   response ([170]34). The inability to identify distinct molecular
   endotypes of fibrotic lung diseases not only hampers personalized
   patient management but has also likely contributed to a string of
   clinical trials with negative outcomes as specific patient subgroups
   that may respond to specific treatments cannot be identified and
   recruited prospectively ([171]35). Molecular endotyping of IPF through
   blood and BAL biomarkers ([172]36–[173]39) shows promise but is
   constrained by its susceptibility to systemic influences beyond the
   lungs in case of blood biomarkers and the invasive nature of BAL. To
   overcome these limitations, molecular imaging has gained increasing
   attention for noninvasive monitoring of key pathogenic drivers of lung
   inflammation-fibrosis axis ([174]15, [175]40).

   Our comprehensive analysis of lung immune cells using large scRNA-seq
   datasets suggested that CMKLR1 expression is a hallmark of nascent
   MDMφ, enriched in patients with fibrotic ILDs; hence, it represents a
   promising biomarker for noninvasive imaging of the lung
   inflammation-fibrosis axis. We previously described [^64Cu]NODAGA-CG34
   as a CMKLR1-targeting peptidomimetic tracer with several favorable
   properties, including facile radiolabeling with a high yield and molar
   activity, plasma stability, low plasma protein binding, and rapid
   clearance ([176]16). Although CG34-derived peptides exhibit lower
   affinities to CMKLR1 (K[d] ~ 200 nM) compared to chemerin ([177]16),
   here, we demonstrated their ability to induce receptor-mediated
   internalization similar to the natural chemerin-derived peptides
   ([178]41), which explains the high accumulation of [^64Cu]NODAGA-CG34
   in inflamed lungs. We demonstrated that CG34-specific binding is
   restricted to CMKLR1 without detectable binding to the other two
   receptors of chemerin, namely, CCRL2 and GPR1.

   Molecular imaging strategies often encounter the limitation of
   targeting a single biomarker, which may not yield the same degree of
   specificity attained through multiparametric invasive analyses, such as
   immunoprofiling by flow cytometry or scRNA-seq. Recognizing this
   limitation, we determined the accuracy of our CMKLR1 targeting approach
   in detecting MDMφ by investigating the uptake of a
   fluorochrome-conjugated analog of [^64Cu]NODAGA-CG34 by different
   immune cells at various stages of bleomycin-induced lung injury. MDMφ
   were the primary contributors to increased 6CF-CG34 uptake during the
   inflammatory phase of lung injury, accounting for ~78 and ~86% of the
   total uptake by the main CMKLR1-expressing cells, i.e.,
   monocytes/macrophages and natural killer cells, at weeks 1 and 2
   postbleomycin, respectively. Conversely, tissue-resident alveolar
   macrophages, monocytes, neutrophils, and lymphocytes showed negligible
   6CF-CG34 uptake. Interstitial macrophages, on the other hand, exhibited
   a transient increase in 6CF-CG34 uptake during the pneumonitis phase of
   lung injury, peaking at week 1 (accounting for ~11% of the uptake) and
   declining by week 2 (~3% of the uptake). This observation may suggest
   CMKLR1 expression during the intermediary differentiation stage of
   MDMφ, as interstitial macrophages precede the differentiation of
   monocytes into alveolar macrophages ([179]42), consistent with our
   scRNA-seq analysis revealing a transient CMKLR1 expression in human
   lungs in their trajectory toward SPP1-Mφ. However, alternative
   explanations, such as the induction of CMKLR1 by resident interstitial
   macrophages, warrant further investigation through fate mapping
   studies. Consistent with our previous report ([180]16),
   CMKLR1-expressing natural killer cells contributed significantly to the
   minimal baseline 6CF-CG34 uptake in noninflamed lungs, although their
   contribution to total uptake during lung injury was overshadowed by
   that of MDMφ.

   [^64Cu]NODAGA-CG34 PET at various time points after bleomycin
   administration closely mirrored the temporal pattern of the
   accumulation of CMKLR1-expressing MDMφ observed by flow cytometry.
   Specifically, we noted an increased [^64Cu]NODAGA-CG34 uptake at 1 and
   2 weeks after bleomycin treatment, which returned to near-baseline
   levels by week 4. Despite challenges inherent to PET imaging of lungs,
   such as respiratory motion ([181]43), we achieved accurate noninvasive
   quantification of lung uptake of [^64Cu]NODAGA-CG34 as evidenced by
   strong correlation between PET-derived and γ-counting–derived
   measurements. We validated the prognostic value of CMKRL1-targeted PET
   by establishing that [^64Cu]NODAGA-CG34 uptake during the early stage
   of bleomycin-induced injury predicts the extent and spatial
   distribution of future pulmonary fibrosis. Together, our findings
   emphasize the feasibility of CMKLR1-targeted PET as a quantitative
   approach to detect the spatiotemporal pattern of MDMφ accumulation
   throughout different stages of fibrotic lung injury for noninvasive
   monitoring of the lung inflammation-fibrosis axis.

   To confirm the translational relevance of CMKLR1 as a biomarker in
   fibrotic lung diseases, we demonstrated that high CMKLR1 expression by
   BAL cells is associated with a distinct IPF endotype with a
   significantly high mortality, in particular in the
   short-to-intermediate timeframe, and enrichment of inflammatory and ECM
   remodeling pathways. This information may have major clinical
   implications, such as allowing for prioritizing lung transplantation or
   implementing more aggressive therapies in patients considered at risk
   of rapid deterioration. Moreover, identifying patients with
   inflammatory endotypes of fibrotic lung diseases may improve the design
   and outcome of clinical trials by allowing for recruitment of patients
   who are more likely to benefit from anti-inflammatory treatments and
   monitoring their therapeutic response. Last, our exploratory
   histological analysis of lung specimens suggests that CMKLR1 is a
   relevant inflammatory biomarker in a wide range of fibrotic lung
   diseases in addition to IPF, such as sarcoidosis, scleroderma,
   rheumatoid arthritis, and coal worker pneumoconiosis. However, to
   establish disease-specific patterns of CMKLR1 expression in these
   conditions, further investigations involving larger sample sizes are
   warranted.

   This study has several limitations. There are relatively few
   preclinical models of pulmonary fibrosis and none entirely replicates
   the pathological complexities of the diverse diseases, which lead to
   lung fibrosis in humans ([182]44). The widely used bleomycin-induced
   lung fibrosis model has been extensively validated and recapitulates
   several key pathological features observed in clinical disease,
   including the release of similar cytokines and growth factors as well
   as activation of relevant signaling pathways ([183]6, [184]30,
   [185]45). However, the single bleomycin injection protocol does not
   replicate the progressive nature and spatiotemporal heterogeneity
   characteristic of usual interstitial pneumonia, the pathological and
   imaging hallmarks of IPF. In contrast, repetitive intratracheal
   administration of bleomycin, as a more recently developed alternative
   animal model, more closely recapitulates the temporal heterogeneity and
   long-term progression of pulmonary fibrosis ([186]46, [187]47). The
   validation of [^64Cu]NODAGA-CG34 PET in the repetitive bleomycin
   instillation model may provide further biological insights and aid in
   distinguishing the inflammatory and ECM remodeling components of
   fibrotic lung injuries. CMKLR1-targeted PET may also serve as a tool
   for mechanistic studies, shedding light on the role of accumulated MDMφ
   in sustaining inflammation beyond the resolution of acute injury and
   driving progressive lung fibrosis. Another limitation of our study is
   that our analysis of the prognostic role of CMKLR1 expression was
   restricted to patients with IPF and used a bulk transcriptomics
   approach as scRNA-seq analysis was not performed in the original study
   ([188]32). While pulmonary fibrosis arises from various underlying
   etiologies, we focused on IPF as the prototypical and most common
   fibrotic ILD. In addition, we are not aware of other publicly available
   BAL transcriptomics dataset with longitudinal follow up in other
   fibrotic lung diseases, which limits our ability to explore the
   prognostic potential of CMKLR1 in other conditions. Another limitation
   of this study is that BAL sampling may underestimate the extent of
   total CMKLR1 expression, which can be detected by PET, as parenchymal
   CMKLR1^+ macrophages ([189]16) are not efficiently recovered by this
   technique. To address this limitation, we performed an exploratory
   study to determine CMKLR1 expression in lung tissues from patients with
   various causes of pulmonary fibrosis. Our findings demonstrated that
   CMKLR1 is widely expressed in the lungs across these etiologies.
   Furthermore, the current study was focused to investigate the promise
   of CMKLR1 as a biomarker of lung inflammation-fibrosis axis, and
   potential mechanistic contributions of CMKLR1 in driving adverse ECM
   remodeling and fibrosis require further studies. Last, because of the
   lack of Food and Drug Administration approval for this tracer, as an
   Investigational New Drug, we were unable to perform PET studies in
   patients. Nonetheless, the promising results obtained from
   [^64Cu]NODAGA-CG34 in preclinical models of lung diseases and the high
   affinity of CG34 to human CMKLR1 along with the clinical relevance of
   CMKLR1 as a biomarker of MDMφ in patients with COVID-19 ([190]16) and
   IPF will lay a solid foundation for future Investigational New Drug
   application to enable the first-in-human use of this tracer in fibrotic
   lung diseases.

   In summary, our study highlights the potential of CMKLR1-targeted PET
   as a valuable tool for noninvasive monitoring of the lung
   inflammation-fibrosis axis through the detection of MDMφ accumulation.
   By providing valuable insights into the molecular processes underlying
   the pathogenesis of the lung inflammation-fibrosis axis, this approach
   holds promise for endotyping of fibrotic lung diseases and may offer
   opportunities for improved prognostication and optimization of
   individualized management strategies in a wide range of inflammatory
   and fibrotic lung diseases. It is noteworthy that MDMφ recruitment
   occurs in various inflammatory lung diseases, such as ARDS ([191]16),
   pneumonia/pneumonitis ([192]48), COPD ([193]49), and cystic fibrosis
   ([194]50), and their pathological effects depend on the context of the
   underlying disease rather than necessarily driving fibrosis.
   Consequently, the application of CMKLR1-targeted PET may extend beyond
   pulmonary fibrosis to various inflammatory lung diseases; hence, it is
   imperative to consider the unique context of the underlying disease for
   a better understanding of the implications of the uptake of
   CMKLR1-targeting tracers.

MATERIALS AND METHODS

Chemicals and reagents

   The Supplementary Materials provide the list of major chemicals and
   reagents (table S2), plasmids (table S2), flow cytometry reagents
   (table S4), and histology antibodies (table S5).

scRNA-seq data processing of lung immune cells

   Preprocessed data from Habermann et al. ([195]21) was downloaded from
   GEO accession [196]GSE135893. Raw sequencing data from Adams et al.
   ([197]20) was downloaded from [198]GSE136831 and reprocessed; read2 5′
   template-switch oligo contaminants and 3′ poly(A) contaminant sequences
   were trimmed using cutadapt (version 3.1), and trimmed reads were
   mapped to human genome GRCh38 with STARsolo (version 2.7.8.a) using
   Gencode release 38.

Cell type labeling

   Cells from each dataset that were identified as immune in their
   original analyses were isolated for downstream analyses in R (version
   4.0.3) with the package Seurat (version 4.3.0.1). Cells were split into
   two groups based on the dataset and integrated using Seurat’s
   reciprocal principal components analysis (RPCA) implementation. For
   each cell, a “stress score” was estimated using the proportion of
   transcripts, which were Fos/Jun and HSP/DNAJ-family heat-shock
   proteins. Before the integration steps, aforementioned stress genes
   were blacklisted from feature selection and this stress score was
   regressed out during the scaling step performed before each dataset’s
   PCA; all downstream RPCA integration parameters were performed under
   default settings.

   Cell type identities were resolved through an iterative and recursive
   process of integration, dimension reduction, graph embedding,
   high-resolution clustering, and expert curation. This process was
   repeated until all cells could be assigned to categorical groups (cell
   types) with positive and negative distinguishing features that could be
   observed consistently across multiple subjects in both datasets. For
   the final UMAP implementation used, principal components 1 to 16 and 18
   to 23 were used; PCs 17 and 24 were avoided because their loadings were
   heavily influenced by metallothionines. The Seurat RunUMAP
   implementation was run with 15 neighbors, a minimum distance of 0.6,
   and repulsion strength and spread parameters both equal to 1, for 500
   iterations with seed of 7. The proportion of myeloid cells in different
   clusters between IPF and control lungs was compared using the
   Mann-Whitney U test.

Pseudotime analysis

   The group of myeloid cells observed belonging to contiguous UMAP
   feature-space (monocytes, macrophage, and nonplasmacytoid dendritic
   cells) were subsetted for downstream analysis in Python (version
   3.9.16) using the package Scanpy (version 1.9.3). Cell neighbors were
   recalculated with Scanpy’s scanpy.pp.neighbors implementation using the
   prior PCA embeddings but only the first 16 principal components (as
   Scanpy does not support noncontiguous sequences of PCs) with 25
   neighbors per cell with the gaussian distance metric and random state
   of 7. The relatively large neighborhood size was chosen to reduce the
   impact of spurious cell-cell adjacencies. Using recalculated
   neighborhoods, all abundant cell types (monocytes, macrophage, and cDC2
   cells) were reclustered with Scanpy’s louvain implementation with a
   resolution of 3 and a random seed of 7. Granular subclusters were
   subjected to Scanpy’s PAGA (version 1.2) implementation with a
   confidence threshold of 0.65. Intersubpopulation edges pruned via PAGA
   were also pruned in the corresponding cell-cell graph to prevent
   spurious connections in feature space before pseudotime estimations.
   Pseudotime estimates for each trajectory were performed via the Scanpy
   external wrapper for Palantir (version 1.0.1) using the pruned
   cell-level adjacency matrix, 15 diffusions components, and 1000
   waypoints per trajectory.

   Pseudotime estimation results were then subjected to downstream
   analysis in R (this time, version 4.3.0) using the package tradeSeq
   (version 1.13.01) to fit GAMs of variable genes to each pseudotime
   ordering trajectory. To downsample cells without exacerbating biases in
   representation along pseudotime, cells from each trajectory were broken
   into 1001 bins in the pseudotime distance range between 0 and 1 by
   0.001 increments, with a maximum of three cells randomly sampled per
   bin (seed = 7). Among these cells, the top 1000 variable genes were
   identified using Seurat’s (version 4.3.0.1) FindVariableFeatures
   implementation with otherwise default parameters. For tradeSeq’s fitGam
   implementation, both trajectories had GAMs fit in parallel, where cells
   belonging to both trajectories were assigned a 0.5/0.5 probability for
   belonging to either trajectory, with seven knots per model using the
   negative binomial fit, with dataset treated as a random covariate.
   While only downsampled cells and highly variable genes were included in
   the GAM analysis, an offset value to normalize each cell was calculated
   using the log transformed total number of transcripts across all genes.

Secondary analysis of scRNA-seq from IPF and control upper versus lower lobes

   Processed scRNA-seq count data of control and IPF lung tissue described
   in Morse et al. ([199]25) was downloaded from GEO accession
   [200]GSE128033. Cell barcodes with at least 700 transcripts and less
   than 15% of transcripts of mitochondrial origin were used for analysis.
   Immune cells were isolated and cell type labels and approximate UMAP
   coordinates were projected onto them using the Seurat implementation
   MapQuery(), with the integrated immune dataset as reference. Marker
   genes were compared between reference and query datasets to confirm
   label-transfer reliability.

Cell culture

   CHO-K1 cells were grown in complete growth media consisting of Ham’s
   F-12K (Kaighn’s) medium supplemented with 10% fetal bovine serum and 1%
   penicillin/streptomycin. Resident peritoneal cells were obtained from
   C57BL/6J mice, as previously described ([201]16), cultured in RPMI
   medium supplemented with 10% fetal bovine serum and 1%
   penicillin/streptomycin.

Transient transfection

   CHO-K1 cells at 70 to 90% confluency were subjected to transient
   transfection with mouse CMKLR1, mouse GPR1, or mouse CCRL2 plasmids
   using Lipofectamine 3000 following the manufacturer’s instructions.

Stable transfection

   CHO-K1 cells at 70 to 90% confluency were transfected with human CMKLR1
   and G[α15] plasmids using Lipofectamine 3000 following the
   manufacturer’s instructions. Transfected cells were next treated with
   selection antibiotics [geneticin (400 μg/ml) and hygromycin B (800
   μg/ml)] and grown for ~2 weeks. Then, individual cell colonies were
   subcultured and grown until sufficient numbers were available to
   evaluate the expression of the transfected gene by flow cytometry.

Radiolabeling

   NODAGA-CG34 was radiolabeled and characterized as previously reported
   ([202]16). Radio–high-performance liquid chromatography was performed
   to ensure a radiochemical purity of >95% for [^64Cu]NODAGA-CG34 before
   its use in all in vitro and in vivo experiments.

Mouse model of fibrotic lung injury

   Animal experiments were conducted in accordance with a protocol
   approved by the University of Pittsburgh Institutional Animal Care and
   Use Committee. To induce fibrotic lung injury, bleomycin (1.5 U/kg) in
   60 μl of phosphate-buffered saline (PBS) was intratracheally
   administered to C57BL/6J mice (~9 to 12 weeks old, Jackson Laboratory,
   strain no. 000664). Control mice were intratracheally injected with 60
   μl of PBS. Mice were euthanized at 1, 2, or 4 weeks after the induction
   of lung injury or if their weight loss exceeded 25% of the initial body
   weight.

[^64Cu]NODAGA-CG34 PET/CT

   At specified time points following administration of bleomycin or PBS,
   mice were injected intravenously with [^64Cu]NODAGA-CG34 (6.76 ± 0.07
   MBq). To address tracer specificity, a co-injection of
   [^64Cu]NODAGA-CG34 with ~100-fold molar excess of nonradiolabeled
   NODAGA-CG34 was performed in a cohort of mice at 2 weeks after
   bleomycin administration. At 90 min postinjection, static PET (~10 min)
   and low-resolution CT (projections, 180; exposure, 140 ms; rotation,
   180°; voltage, 80 kV; current, 500 μA; and field of view, 78.5 × 100
   mm) were performed using a Siemens Inveon scanner. Kinetic PET/CT from
   ~5 to 90 min postinjection was performed in a cohort of mice using the
   same parameters. High-resolution CT in the longitudinal study was
   performed using the following parameters: projections, 220; exposure,
   800 ms; rotation, 220°; voltage, 60 kV; current, 500 μA; and field of
   view, 30.2 × 31.1 mm.

   Regions of interest were drawn over the left and right lungs of
   attenuation-corrected PET images, and the uptake of [^64Cu]NODAGA-CG34
   in the left and right lungs was averaged and quantified as the mean and
   maximal percentage of injected dose per ml of tissue (%ID/ml[mean] and
   %ID/ml[max]) using IRW software. To quantify the blood pool
   concentration of [^64Cu]NODAGA-CG34 by PET, a spherical region of
   interest was drawn inside of the left ventricle. Biodistribution was
   performed by γ-counting (Wizard^2, PerkinElmer) of harvested organs,
   and data were presented as decay-corrected percentage of injected dose
   per gram tissue (%ID/g). Multiplanar reconstructions were performed
   using software to align PET/CT with autoradiography and histology
   images.

[^64Cu]NODAGA-CG34 autoradiography

   Mice were euthanized following PET/CT, and their lungs were inflated
   with 1 ml of optimal cutting temperature compound (OCT). Subsequently,
   the lungs were embedded in OCT and frozen for cryosectioning (10 μm
   thickness) within ~24 hours of tracer injection. After an overnight
   exposure of tissues to phosphor screens (GE, BAS-IP SR2025 Super
   VivoQuant Resolution), screens were scanned at a resolution of 100 μm
   using a Sapphire Biomolecular Imager, as previously described
   ([203]16).

Mouse lung immunostaining

   After euthanasia, the pulmonary circulation of mice was perfused with
   ~5 ml of PBS through a right ventricular puncture. The lungs were then
   inflated by 1 ml of intratracheal OCT and stored frozen. Cryosections
   of 10 μm thickness were prepared. Tissues were fixed with a cold
   methanol-acetone mixture (1:1). Next, primary antibodies at a dilution
   of 1:200 were applied to the sections and incubated overnight at 4°C.
   Following PBS washes, fluorescently conjugated secondary antibodies
   were added and incubated for 2 hours at 4°C. Slides were mounted with
   ProLong Gold Antifade Mountant containing 4′,6-diamidino-2-phenylindole
   (DAPI) and visualized using an Axio Vert microscope (Zeiss).

Binding/uptake of 6CF-CG34 and 6TAM-CG34 by peritoneal macrophages

   Peritoneal cells were incubated in indicator-free RPMI in the presence
   of increasing concentrations of 6CF-CG34 or 6TAM-CG34, both with and
   without co-incubation with unlabeled Chem[145–157] (high-affinity
   CMKLR1 ligand derived from the amino acids 145 to 157 of chemerin;
   peptide sequence, P-H-S-F-Y-F-P-G-Q-F-A-F-S-COOH; 2.5 μM) ([204]16),
   for 1 hour at 37°C. After a single wash with PBS, flow cytometry was
   performed using LSR II Flow Cytometer (BD Biosciences) and analyzed by
   FlowJo software version 10.7.2 (BD Biosciences). Total, nonspecific,
   and specific saturation binding curves were generated using GraphPad
   Prism software.

Competitive binding/uptake assay of 6CF-CG34 or 6TAM-CG34 versus NODAGA-CG34
in peritoneal macrophages

   Peritoneal cells were incubated in indicator-free RPMI in the presence
   of 6CF-CG34 or 6TAM-CG34 (50 nM) with increasing concentrations of
   NODAGA-CG34 (10^−10 M to 10^−5 M) for 1 hour at 37°C. After a single
   wash with PBS, flow cytometry was performed using LSR II Flow Cytometer
   and analyzed by FlowJo software. Competitive binding curves were
   generated using GraphPad Prism Software.

Binding/uptake of fluorescently tagged CG34 peptides by CHO-K1 cells
transiently expressing mouse or human CMKLR1, mouse GPR1, or mouse CCRL2

   CHO-K1 cells transiently expressing mouse CMKLR1 (mCMKLR1), human
   CMKLR1 (hCMKLR1), mouse GPR1 (mGPR1), or mouse CCRL2 (mCCRL2) were
   incubated in indicator-free RPMI in the presence of increasing
   concentrations of 6CF-CG34 or 6TAM-CG34 with or without co-incubation
   with unlabeled CG34 (2.5 μM) for 1 hour at 37°C. After a single wash
   with PBS, flow cytometry was performed using a FACSCalibur or LSR II
   Flow Cytometer (BD Biosciences) and analyzed by FlowJo software. Total,
   nonspecific, and specific saturation binding curves were generated
   using GraphPad Prism Software.

Internalization of 6TAM-CG34 by mouse CMKLR1

   CHO-K1 cells transiently expressing mCMKLR1 or nontransfected cells
   grown on glass coverslips coated with fibronectin were incubated in
   indicator-free RPMI with 6TAM-CG34 (250 nM) with or without
   co-incubation with CG34 (10 μM) for 1 hour at 37°C. After a single wash
   with PBS, cells were incubated with Alexa Fluor 647–conjugated wheat
   germ agglutinin (2 μg/ml) for 20 min at 4°C to label the cell membrane.
   Subsequently, cells were fixed with formalin and coverslips were
   mounted onto glass slides with ProLong Gold Antifade Mountant
   containing DAPI and photographed with an Axio Vert microscope.

Flow cytometric immunophenotyping of murine lung cells

   Following euthanasia of bleomycin- or PBS-treated mice, the lungs were
   perfused with PBS. Mechanical dissociation of the lungs was performed
   using a gentleMACS Dissociator (Miltenyi) as described previously
   ([205]16). The cells were then divided into separate tubes and
   incubated in indicator-free RPMI in the absence or presence of 50 nM
   6CF-CG34, with or without co-incubation with 10 μM Chem[145–157], for 1
   hour at 37°C. After a PBS wash, nonspecific binding was blocked using
   1% bovine serum albumin and Fc block (2 μl per sample) before
   incubation with a mixture of antibodies (0.5 μl per sample) and DAPI (5
   μl of 1 μg/ml per sample) for 0.5 hour at 4°C. Last, cells were washed
   and fixed with 4% formalin for 0.5 hour at room temperature. Flow
   cytometry was performed using an LSR II Flow Cytometer and analyzed by
   FlowJo software. We defined distinct lung macrophages based on their
   differential expression of three surface markers, i.e., alveolar
   macrophages (CD11b^lo/CD11c^hi/SiglecF^hi), interstitial macrophages
   (CD11b^hi/CD11c^lo/SiglecF^lo), and MDMφ
   (CD11b^hi/CD11c^hi/SiglecF^int), as previously confirmed by lineage
   tracing studies ([206]6, [207]7). The number of 6CF-CG34 molecules
   bound to/internalized by cells was calculated based on a standard curve
   generated using the FITC Easy Calibration Kit (Spherotech) according to
   the manufacturer’s instructions. Because of the low recovery by
   mechanical dissociation, dendritic cells were excluded from the
   subsequent analysis ([208]16).

Hydroxyproline assay

   The lungs of euthanized mice were frozen on dry ice and stored at
   −80°C. Hydroxyproline contents of the left lungs were measured using a
   commercially available kit following the manufacturer’s instruction
   (Sigma-Aldrich).

Transcriptomics analysis of BAL cells in patients with IPF

   A secondary analysis of a previously published ([209]32) and publicly
   available BAL transcriptome dataset ([210]GSE70867) in patients with
   IPF (n = 176) and control participants (n = 20) was conducted. The
   characteristics of the patients and controls as well as the methodology
   for microarray bulk mRNA transcriptomics were described in the original
   paper ([211]32). Patients were recruited from three sites: Freiburg,
   Germany (n = 62); Leuven, Belgium (n = 64); and Siena, Italy (n = 50).
   BAL samples were collected before the initiation of antifibrotic or
   anti-inflammatory medications. To account for variations across study
   sites, batch effect adjustment was performed using the ComBat function
   [in SVA R package version 3.46.0 ([212]51, [213]52), RStudio
   statistical system, version R 4.2.2]. Differential gene expression
   analysis was conducted (using limma package version 3.54.2) ([214]53)
   between patients with below-median and higher-than-median CMKLR1
   expression in BAL cells. Next, we conducted a pathway enrichment
   analysis using the enrichKEGG function from the clusterProfiler R
   package (version 4.7.1.003). All genes were mapped to Entrez IDs. Our
   pathway analysis was limited to genes that were significantly
   overexpressed (adjusted P < 0.01 by Benjamini-Hochberg procedure) in
   patients with higher-than-median versus below-median CMKLR1 expression.

   To determine the prognostic role of CMKLR1 expression in predicting
   survival of patients with IPF, we performed Kaplan-Meier analysis after
   stratifying patients into tertiles of CMKLR1 expression. Next, we
   studied whether CMKLR1 expression level is an independent predictor of
   survival using multivariable Cox proportional hazards regression
   analysis (SPSS) by adjusting for potential confounders, age, sex, study
   sites, and GAP score. Further, to evaluate whether CMKLR1 expression
   provides incremental prognostic information compared to the risk
   stratification achieved by the GAP index, we performed Kaplan-Meier
   analysis after stratifying patients into low versus high GAP index
   groups (0 to 4 versus 5 to 8, respectively). Time-varying ROC curves
   were plotted to compare the potential of CMKLR1 expression and GAP
   index in predicting mortality over the course of 1 to 4 years after
   BAL. Kaplan-Meier analyses were also performed after stratifying
   patients into tertiles of CCR2 and SPP1 expression.

Human lung immunostaining

   Deidentified paraffin-embedded lung tissues from patients with
   different types of fibrotic lung diseases and controls (table S1) were
   obtained from the University of Pittsburgh Biorepository Core. Signed
   informed consents were received before organ procurement for
   participation in the biorepository and secondary research use under a
   protocol approved by the University of Pittsburgh Institutional Review
   Board. A determination of Not Human Subject research was made for this
   specific study (no. 23070112) by the University of Pittsburgh
   Institutional Review Board, which involved secondary analysis of
   deidentified lung tissues sections that were not collected specifically
   for this study. Immunostaining was performed after antigen retrieval
   achieved by heating 5-μm-thick tissues in sodium citrate buffer (10 mM)
   to 90°C for 20 min. The primary antibodies, anti-CMKLR1 (1:2000
   dilution) and anti-CD45 (1:500 dilution), were applied and incubated at
   4°C overnight. Following PBS washes, secondary antibodies (1:200
   dilution) were applied to the tissues and incubated for 120 min at 4°C.
   Slides were mounted with ProLong Gold Antifade Mountant containing DAPI
   and photographed using an Axio Vert microscope.

Statistical analysis

   Statistical analysis of human lung scRNA-seq and BAL transcriptomics
   data are described in the relevant section. Statistical analyses of
   murine data were performed using Prism 9 (GraphPad). The data are shown
   as the mean ± SEM. A Student’s t test was performed to compare the mean
   values between two groups. One-way analysis of variance (ANOVA),
   followed by Fisher’s exact post hoc test, was used to compare the mean
   values in >2 groups. Correlations between variables were analyzed by
   the Pearson test. Statistical significance was considered as P < 0.05.

Acknowledgments