Detection of Pulmonary Fibrosis with a Collagen-Mimetic Peptide

Abstract

Idiopathic pulmonary fibrosis (IPF) is a disease of unknown etiology that is characterized by excessive deposition and abnormal remodeling of collagen. IPF has a mean survival time of only 2–5 years from diagnosis, creating a need to detect IPF at an earlier stage when treatments might be more effective. We sought to develop a minimally invasive probe that could detect molecular changes in IPF-associated collagen. Here, we describe the design, synthesis, and performance of [⁶⁸Ga]Ga·DOTA–CMP, which comprises a positron-emitting radioisotope linked to a collagen-mimetic peptide (CMP). This peptide mimics the natural structure of collagen and detects irregular collagen matrices by annealing to damaged collagen triple helices. We assessed the ability of the peptide to detect aberrant lung collagen selectively in a bleomycin-induced mouse model of pulmonary fibrosis using positron emission tomography (PET). [⁶⁸Ga]Ga·DOTA–CMP PET demonstrated higher and selective uptake in a fibrotic mouse lung compared to controls, minimal background signal in adjacent organs, and rapid clearance via the renal system. These studies suggest that [⁶⁸Ga]Ga·DOTA–CMP identifies fibrotic lungs and could be useful in the early diagnosis of IPF.

Graphical Abstract

Collagen is the most abundant protein in the human body.¹ This abundance avails the extraordinary stability conferred by its hierarchical structure. Collagen is a right-handed triple helix consisting of three left-handed polyproline type-II (PPII) helices.^1,2 Its amino acid sequence has a repeating triplet, XaaYaaGly, where Xaa is most commonly (2S)-proline (Pro) and Yaa is most commonly (2S,4R)-4-hydroxyproline (Hyp).³ The high abundance of Pro and Hyp residues preorganize collagen strands for helix formation, whereas the small glycine (Gly) residue fits into the core of the triple helix.⁴

In healthy tissue, collagen helices undergo extensive proteolytic remodeling for tissue maintenance.⁵ When these processes are dysregulated, excessive levels of proteolysis-destabilized collagen triple helices are produced that can denature at biological temperature.⁶ This damaged collagen can be targeted with collagen-mimetic peptides (CMPs), which mimic the natural structure of collagen and can anneal to dissociated collagen (Figure 1).^7,8 Thus, CMPs can anchor pendant molecules to tissues with damaged or abnormal collagen.^8-11

Figure 1.

Conceptual representation of a collagen-mimetic peptide annealed to a damaged collagen triple helix. A pendant moiety, “X”, becomes anchored in the damaged collagen.

The extracellular matrix is damaged or abnormal in a variety of fibroproliferative diseases. For example, pulmonary fibrosis, which is a pathology associated with a range of lung diseases, is characterized by dysregulated matrix metalloproteinase production.¹² Idiopathic pulmonary fibrosis (IPF) is the prototypic form of pulmonary fibrosis and is associated with particularly severe symptoms.¹³

We reasoned that CMPs could be noninvasive probes for detecting pulmonary fibrosis. We anticipated two possible means by which collagen could lose integrity. First, rapid collagen production and abnormal collagen remodeling in IPF lungs could lead to an incorrectly remodeled collagen matrix. During fibrosis, collagen triple helices might not be adequately interdigitated for fibril formation, yielding less stable triple helices that are more likely to possess dissociated domains where CMPs can anneal.¹⁴ This modality is supported by the correlation of an IPF phenotype with diminished collagen ordering within a tissue.¹⁵ Tropocollagen triple helices have lower thermostability than do helices incorporated into higher-order structures and are unstable at 37 °C,^6,14 thus increasing the propensity for IPF triple helices to denature in a mammal and provide binding sites for CMPs. Second, matrix-metalloproteinases are upregulated in active pulmonary fibrosis and in the corresponding mouse model.¹⁶ These enzymes cleave strands in collagen triple helices, again providing binding sites for CMPs.¹²

The current paradigm for IPF diagnosis is high-resolution computed tomography (HRCT). The median survival of IPF ranges from 2–5 years from diagnosis,¹⁷ emphasizing the need for early detection of IPF when treatments might be more effective. Clinical trial data with two approved antifibrotic therapies showed a benefit in patients with more preserved lung function and imply greater benefit if disease could be treated earlier, which would require early detection.^18,19 We sought to leverage positron emission tomography (PET) to achieve this goal, as this modality is known for its high sensitivity and ability to detect molecular markers. Developing a collagen-specific PET tracer could serve as a complementary tool to conventional HRCT. For example PET could be valuable in the early detection of IPF, especially in patients at high risk for the disease, such as those with a family history of IPF²⁰ or interstitial lung abnormalities.²¹ Moreover, in cases where HRCT cannot distinguish between fibrosis and nonfibrotic inflammation, a biopsy is required to identify signs of fibrosis histologically.²² Unfortunately, biopsy carries risks of morbidity and mortality and only samples a tiny fraction of the lung.²³ The direct visualization of damaged collagen as a molecular signature of fibrosis has the potential to identify fibrosis throughout the entire lung. Other collagen-targeted PET probes have been developed for minimally invasive pulmonary fibrosis detection;^14,24-26 however, these probes generally bind to intact rather than damaged collagen.

We sought to enhance existing IPF diagnostic techniques by designing a novel, minimally invasive probe to detect molecular changes in IPF-associated collagen. Design criteria led us to [⁶⁸Ga]Ga·DOTA–CMP, which is detectable by PET/MR, an imaging modality that delivers a lower radiation dose than PET/CT while supplying multifaceted MR details on soft tissue anatomy. [⁶⁸Ga]Ga·DOTA–CMP contains the peptide (Gly-Ser)₂-Gly-(flp-Hyp-Gly)₇, where flp refers to (2S,4S)-4-fluoroproline. This peptide can anneal to damaged collagen triple helices (Figure 2A).^8,27 Importantly, this peptide resists homotrimer formation, a process that competes deleteriously with forming a triple helix with damaged collagen.^8,27,28 That resistance is due to a steric clash between a 4S-substituted and a 4R-substituted proline residue in each register of a triple helix.^29,30 The nonnatural residue flp differs from the natural residue Pro only by the replacement of a single H atom with F, and the H-flp-OH amino acid is not toxic to human cells.⁸

Figure 2.

(a) Structure of [⁶⁸Ga]Ga·DOTA–CMP. The CMP segment is depicted as a red ribbon in Figure 1. (b) Structure of [⁶⁸Ga]Ga·DOTA–CI.

As a control, we synthesized a compositional isomer (CI), [⁶⁸Ga]Ga·DOTA–CI, which contains the peptide (Gly-Ser)₂-Gly-(Hyp-flp-Gly)₇ (Figure 2B). The Hyp⟷flp sequence permutation interferes with the stereochemical requirements for forming a collagen helix.^31,32 Accordingly, this peptide cannot anneal to damaged collagen, despite its amino acid composition.⁸

The N-terminal Gly-Ser-Gly-Ser-Gly sequence serves as a flexible, hydrophilic linker for conjugation to the radioisotope chelator, 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTA). We selected DOTA because of the high stability of its complex with ⁶⁸Ga, a positron-emitting radioisotope that does not require a cyclotron for generation.^{33,34 68}Ga has a half-life of 68 min, providing an ideal half-life for PET imaging while requiring a dose that is low enough to mitigate total radiation exposure.³⁵ This isotope decays into ⁶⁸Zn, which is an essential trace element.^{36 68}Ga has been widely applied in the clinical community for PET, facilitating clinical translation.^37,38

The two peptides were synthesized via automated microwave-assisted Fmoc-mediated solid-phase peptide synthesis (Scheme S1) and purified by reversed-phase high-performance liquid chromatography (RP-HPLC). The purified peptides were conjugated to NHS-ester-activated DOTA at the N-terminus. Following conjugation, the DOTA–peptide conjugates were purified by preparative RP-HPLC. Their purities were verified with analytical RP-HPLC (Figure S1) and matrix-assisted laser desorption/ionization–time-of-flight mass spectrometry (Figure S2).

The structure of DOTA–CMP was assessed by circular dichroism (CD) spectroscopy. A solution of DOTA–CMP displayed a shallow peak at 225 nm and no cooperative denaturation with an increasing temperature (Figure S3). These data are consistent with signatures of PPII structures that do not form triple helices and confirm that DOTA–CMP remains to collagen strands, we recorded CD spectra of a 1:1 mixture of DOTA–CMP and (ProProGly)₇. (ProProGly)₇ was used as a proxy for damaged collagen because it bears the canonical XaaYaaGly motif but is too short to form a homotrimer, avoiding a confounding CD signal in the mixture.^8,29 Additionally, (ProProGly)₇ does not have a 4R-substituted proline residue, thereby avoiding steric clashes with the flp residues of DOTA–CMP.²⁹ The CD spectra of the 1:1 mixture of DOTA–CMP:(ProProGly)₇ exhibited a strong peak at 225 nm, cooperative denaturation with increasing temperature, and a T_m value of 34 °C (Figure S3). These data suggest that the addition of DOTA to the N terminus of the CMP did not impede its ability to form heterotrimers with collagen.

Our strategy relies on the CMP being stable in sera and plasma. We assessed this stability in two ways. First, we incubated 5(6)-TAMRA–(Gly-Ser)₂-Gly-(flp-Hyp-Gly)₇-NH₂ in sera. During a 48-h time course, we detected no degradation in human serum and only slight degradation in mouse serum (Figure S4). Then, we radiolabeled DOTA–CMP with ⁶⁴Cu by adapting a known procedure.³⁹ (The 12.7 h half-life of ⁶⁴Cu provides a longer time course for this analysis than does ⁶⁸Ga.) We observed no degradation or proteolysis of ⁶⁴Cu·DOTA–CMP over 21 h in human plasma, as analyzed with instant thin-layer chromatography (Figure S5). This high stability in serum and plasma is consistent with that of other CMPs.⁴⁰

Next, we radiolabeled DOTA–CMP and DOTA–CI with ⁶⁸Ga,³⁹ giving the desired product with ≥90% radiochemical yield as determined by radio-HPLC (Figure S6). We assessed the ability of the two ⁶⁸Ga-labeled conjugates to detect IPF in vivo by using a validated mouse model of IPF.^41,42 To create the model, we injured male C57BL/6 mice with a single transtracheal instillation of bleomycin in phosphate-buffered saline (PBS) (Figure S7A).¹⁴ Fourteen days later, the mice developed pulmonary fibrosis, which was confirmed by histopathological analyses. Those analyses revealed excessive interstitial deposition of collagen as demonstrated by staining collagen with Picrosirius red as well as the destruction of lung architecture (Figure S8).^14,42 As a control, sham animals, which received transtracheally administered PBS, were healthy and showed no signs of pulmonary disease (Figure S6), in line with previously reported results.¹⁴ Mice were injected in their tail vein with [⁶⁸Ga]Ga·DOTA–CMP or [⁶⁸Ga]Ga·DOTA–CI (50–150 μCi) and imaged with 4.7 T PET/MR. After imaging, the mice were sacrificed for ex vivo analysis 90 min after administration of the probes. PET imaging showed that [⁶⁸Ga]Ga·DOTA–CMP was not retained in the lungs of sham mice and displayed a nearly undetectable signal in all satellite organs (Figure 3). Background signal was observed in the kidneys and bladder, which are involved in renal excretion of the probe following its rapid elimination from the blood (estimated t_1/2 = 6.3 min). In comparison to sham mice, [⁶⁸Ga]Ga·DOTA–CMP was taken up in the lungs of IPF mouse models (Figures 3, S9, and S10). Comparable biodistribution was observed in satellite organs. In vivo quantification of the decay-corrected PET signal in the lungs averaged over 50–60 min revealed that [⁶⁸Ga]Ga·DOTA–CMP had a 4.9-fold higher lung uptake in IPF mouse models than sham mice, P < 0.001 (Figure 4A). Additionally, the ratio of specific uptake in the right lung to nonspecific uptake in muscle (RLMR) was compared in sham and bleomycin-injured mice injected with [⁶⁸Ga]Ga·DOTA–CMP. The RLMR was also found to be 2.6-fold higher for bleomycin-injured mice compared to sham mice (Figure 4B). Following imaging, mice were sacrificed and dissected, and the uptake of probe in individual excised organs was quantified with a gamma counter. Minimal background signal was observed in satellite organs, excluding those required for expected renal excretion (Figure S11).

Figure 3.

PET and fused PET–MR images of sham and bleomycin-injured mice 50–60 min after tail-vein injection with [⁶⁸Ga]Ga·DOTA–CMP or [⁶⁸Ga]Ga·DOTA–CI (A) Representative coronal maximum-in-projection (MIP). (B) Representative axial fused PET (color scale) – MR (gray scale) images of the lungs and heart.

Figure 4.

Radiation signal in organs from sham and bleomycin-injured mice injected with [⁶⁸Ga]Ga·DOTA–CMP or [⁶⁸Ga]Ga·DOTA–CI. (A) In vivo %ID/cc values in right lungs 50–60 min post-injection as measured by PET. (B) In vivo right lung:muscle ratios (RLMRs) 50–60 min post-injection as measured by PET. In panels A and B, n = 6 for [⁶⁸Ga]Ga·DOTA–CMP + sham and n = 5 for [⁶⁸Ga]Ga·DOTA–CMP + bleomycin. (C) and (D) show data from a paired study, in which mice were injected with [⁶⁸Ga]Ga·DOTA–CMP and imaged, then 24 h later, injected with [⁶⁸Ga]Ga·DOTA–CI and imaged. (C) In vivo %ID/cc values in right lungs 50–60 min post-injection as measured by PET. Connecting lines indicate associated data points from individual animals. (D) In vivo RLMRs 50–60 min post-injection as measured by PET. Connecting lines indicate associated data points from individual animals. (E) Ex vivo %ID/lung left uptake 90 min post-injection. In panel E, n = 6 for [⁶⁸Ga]Ga·DOTA–CMP + sham, n = 13 for [⁶⁸Ga]Ga·DOTA–CI + bleomycin, and n = 5 for [⁶⁸Ga]Ga·DOTA–CMP + bleomycin. An unpaired t-test was performed on the data in panels A and B. A paired t-test was performed on the data in panels C and D. A one-way ANOVA with a post-hoc Tukey test was performed on the data in panel E. Values are the mean ± SE. *P < 0.05, **P < 0.01, ***P < 0.001.

To examine the specificity of [⁶⁸Ga]Ga·DOTA–CMP further, we performed a paired study of mice injected with both [⁶⁸Ga]Ga·DOTA–CMP and the control probe, [⁶⁸Ga]Ga·DOTA–CI (Figure S7B). For this study, mice were injected with [⁶⁸Ga]Ga·DOTA–CMP and imaged. After 24 h, ⁶⁸Ga had decayed to background levels, and the same mice were injected with [⁶⁸Ga]Ga·DOTA–CI and imaged. As with [⁶⁸Ga]Ga·DOTA–CMP in sham mice, the control peptide [⁶⁸Ga]Ga·DOTA–CI was not taken up in the lungs of IPF mouse models. In vivo quantification of the decay-corrected PET signal in the lungs averaged over 50–60 min revealed that [⁶⁸Ga]Ga·DOTA–CMP had a significantly (3.6-fold) higher lung uptake in IPF mouse models compared to [⁶⁸Ga]Ga·DOTA–CI in the same mouse model (Figure 4C). We also found that [⁶⁸Ga]Ga·DOTA–CMP had a significantly (1.7-fold) higher RLMR, supporting the selectivity of CMPs for collagen in lungs afflicted with bleomycin-induced pulmonary fibrosis (Figure 4D).

To further support the selectivity of our peptide for pulmonary fibrotic tissue and not other organs, we integrated the in vivo PET signal of the kidneys over the time of the study, that is, from 0 to 60 min after injection into bleomycin-injured mice that had been treated with [⁶⁸Ga]Ga·DOTA–CMP and [⁶⁸Ga]Ga·DOTA–CI on consecutive days (Figure S7B). We did not observe a significant difference in organ retention (P = 0.0878; Figure S12), suggesting that the higher kidney PET signal in the bleomycin-injured animal in Figure 3 compared to the sham mouse is simply due to slower renal clearance in the sick mouse and is not an effect of [⁶⁸Ga]Ga·DOTA–CMP binding selectively to a target in the kidneys of mice with localized lung injury.

To better reflect the trends observed in tissues, we compared the total ex vivo probe uptake in each lung (Figure 4E). Sham mice injected with [⁶⁸Ga]Ga·DOTA–CMP and IPF mouse models injected with [⁶⁸Ga]Ga·DOTA–CI did not differ significantly in probe uptake and biodistribution. This resemblance suggests that the lung enrichment observed in bleomycin-injured mice treated with the CMP was due to CMP binding, rather than a general lag in the clearance of the probe. In contrast, the lung uptake of [⁶⁸Ga]Ga·DOTA–CMP was 11- to 12-fold higher in IPF mouse models than sham mice and 3-fold higher than the uptake of [⁶⁸Ga]Ga·DOTA–CI in IPF mouse models.

Finally, we estimated the mass of collagen in the excised mouse lungs by quantifying Hyp, which is a common residue in collagen but not other proteins.⁴³ We then compared the mass of H-Hyp-OH found in each lung to the ex vivo lung uptake of [⁶⁸Ga]Ga·DOTA–CMP. We found that the uptake of [⁶⁸Ga]Ga·DOTA–CMP had little correlation with the lung Hyp content (R² = 0.33) (Figure S13), suggesting that this probe did not merely bind to overexpressed collagen in bleomycin-injured mice. These data are consistent with the binding of [⁶⁸Ga]Ga·DOTA–CMP to a specific biomarker of pulmonary fibrosis—damaged collagen triple helices (Figure 1).

In conclusion, we described the first use of CMPs as an in vivo probe for detecting pulmonary fibrosis. We find that CMPs report on molecular-level defects in disease-associated collagen, specifically, the presence of non-triple-helical collagen in a fibrotic lung. [⁶⁸Ga]Ga·DOTA–CMP displays specificity and strong uptake in the lungs of a bleomycin-injured IPF mouse model, resulting in up to 11-fold higher uptake in an IPF mouse model compared to controls. The unique mode-of-action of CMPs enables the detection of abnormal collagen production and remodeling, such as those found in pulmonary fibrosis. The specificity of [⁶⁸Ga]Ga·DOTA–CMP in vivo and its ability to selectively detect IPF make CMPs promising candidates for the diagnosis of other fibrotic diseases.