Evaluation of integrin αvβ6 cystine knot PET tracers to detect cancer and idiopathic pulmonary fibrosis

Advances in precision molecular imaging promise to transform our ability to detect, diagnose and treat disease. Here, we describe the engineering and validation of a new cystine knot peptide (knottin) that selectively recognizes human integrin αvβ6 with single-digit nanomolar affinity. We solve its 3D structure by NMR and x-ray crystallography and validate leads with 3 different radiolabels in pre-clinical models of cancer. We evaluate the lead tracer’s safety, biodistribution and pharmacokinetics in healthy human volunteers, and show its ability to detect multiple cancers (pancreatic, cervical and lung) in patients at two study locations. Additionally, we demonstrate that the knottin PET tracers can also detect fibrotic lung disease in idiopathic pulmonary fibrosis patients. Our results indicate that these cystine knot PET tracers may have potential utility in multiple disease states that are associated with upregulation of integrin αvβ6.

T hrough the use of high affinity ligands, positron emission tomography (PET) imaging is an effective way to distinguish between diseased and healthy tissues. Accumulation of PET tracers in tissues provides a way to image the molecular signature of disease, such as cancer by targeting overexpressed cell surface proteins. Aberrant protein expression also occurs in other diseases such as idiopathic pulmonary fibrosis (IPF), a chronic, progressive fibrotic interstitial lung disease (ILD) of unknown cause 1 . Analogous to many cancers, prognosis in IPF is poor with a mean survival of~3-5 years 2 .
Integrin receptors are the focus of extensive efforts aimed at the development of PET tracers [3][4][5][6] . Integrins are a family of proteins that mediate cellular adhesion to extracellular matrix (ECM) proteins. In normal cells, integrins transduce signals from the cell surface to support gene expression of multiple proteins that regulate differentiation, migration, proliferation, and apoptosis 7 .
In certain diseases such as cancer, the expression of some integrins become dysregulated 8 . In a well-studied example, integrin α v β 3 (the angiogenesis integrin) is highly overexpressed on both tumor-neovasculature, and the surface of some human cancer cells 8 . Other integrins, such as α v β 1 and α v β 5 can be biomarkers of specific types of cancer 8 . Similarly, fibrosing diseases are known to express unique subsets of integrins 2,[9][10][11] . Expression expression profiles of different integrins have shown prognostic value, so detecting them in vivo may be useful for diagnosing, managing, or treating disease 8,12,13 . A number of PET tracer and therapeutics, such as the RGD derivatives, that target integrin receptors are currently undergoing clinical trials 5,12 .
A different member of the integrin family, α v β 6 , is overexpressed on the surface of many types of cancer cells 14 . Integrin α v β 6 overexpression has been confirmed in oral squamous cell carcinoma 15 , pancreatic ductal adenocarcinomas (PDAC) 16 , intestinal gastric carcinomas 16 , ovarian cancer 17 , and stage III basal cell carcinoma 18 . Integrin α v β 6 overexpression is a prognostic indicator of reduced survival in colon carcinoma 19 , nonsmall cell lung cancer 20 , cervical squamous carcinoma 21 , and gastric carcinoma 22 . Furthermore, α v β 6 expression is associated with increased liver metastasis of colon cancer 23 . In contrast, its expression is undetectable in many normal tissues including ovary 17 , kidney, lung, skin 15 , and pancreas 16 . In recent studies, immunohistochemical analyses (IHC) revealed robust expression of α v β 6 in PDAC compared to cancers of other systems 14,16 . Expression of α v β 6 was reported to be significantly higher in PDAC compared to chronic pancreatitis, and its expression was observed in tumor-positive lymph nodes 24 . In well differentiated pancreatic tumors, elevated levels of α v β 6 were reported in 100% of the samples (n = 34) by IHC 14,16 . Currently, different research groups, including ours, are developing peptide-based PET tracers to image α v β 6 for cancer detection [25][26][27] .
Integrin α v β 6 has also been the subject of numerous studies as a potential biomarker of fibrotic ILD, which include IPF, nonspecific interstitial pneumonitis (NSIP), chronic hypersensitivity pneumonitis (HP) 2,28 . α v β 6 is a potent activator of transforming growth factor (TGF)β1, which enhances matrix deposition by fibroblasts in wound healing and fibrosis 29,30 . In a recent study, where a number of potential biomarker candidates were screened by tissue IHC, only α v β 6 was found to be statistically associated with a poor prognosis in fibrotic ILDs. Levels of α v β 6 correlate directly and significantly to the risk of death from IPF. A 25month median survival was observed in patients with the highest levels of α v β 6 expression 2 .
Targeted PET ligands created by directed evolution strategies often demonstrate exquisite selectivity and high binding affinity their targets 31,32 . A given PET tracer's pharmacokinetic profile is a function of its biochemical and biophysical characteristics, such as the overall electrostatic charge and molecular weight. Cystine knot peptides (knottins) are small (~4 kDa) peptides characterized by three threaded disulfide bonds arranged in a topological knot; this stabilizing core motif is known as a cystine knot 31,33 . Solvent exposed loops, that can be bioactive, extend from this knotted core. In our studies, we have engineered both high-affinity antigen recognition (K D~1 nM) and pharmacokinetic stabilizers into these loops 26 . Since peptides are generally small, modification of a single amino acid in the framework could have a dramatic effect on overall tracer pharmacokinetics. One advantage of the knottin scaffold is that their backbone residues are highly variable such that they can be fine-tuned to greatly increase tumor uptake or decrease renal uptake in mouse models 26 . In this study, we developed several pharmacokinetically stabilized α v β 6 knottins and evaluated their ability to image disease in living systems as [ 64  Here, we describe the development and validation of knottin PET tracers that selectively recognizes human α v β 6 with singledigit nanomolar affinity. We solve its 3D structure by NMR and x-ray crystallography and validate leads with 3 different radiolabels in pre-clinical models of cancer. We evaluate the lead tracer's safety and performance in healthy human volunteers and show its ability to detect multiple cancers (pancreatic, cervical, and lung). Additionally, we demonstrate the knottin PET tracer's ability to detect fibrotic lung disease in IPF patients. Our studies show the cystine knot PET tracers' potential in multiple disease states associated with elevated α v β 6 .

Results
Development of the lead candidate R 0 1-MG. Three cystine knot peptides (R 0 1, R 0 1-MG, R 0 1-MR) were engineered for stability, and high-affinity molecular recognition of human α v β 6 . Equilibrium binding assays indicated that all variants bind α v β 6 with low single-digit nanomolar binding affinity (K D~1 nM, Supplementary Fig. 1). As a result of comparative imaging studies in vivo, R 0 1-MG emerged as the lead translational candidate (Fig. 1a). It was produced by cGMP solid phase peptide synthesis, and its mass was verified by MALDI-TOF-MS (Supplementary  Table 1). In addition, the three-dimensional (3D) 1 H-NMR solution structure of R 0 1-MG provided evidence that the cystine knot structural element, which defines knottin peptides, was conserved throughout the directed evolution process (Fig. 1b, Table 1). Furthermore, x-ray crystallography studies confirmed a disulfide bond pattern consistent with the cystine knot motif (Fig. 1c, Supplementary Fig. 2, Table 2). Competition binding assays demonstrated that the N-terminus imaging label, located 20 Å from the RTDL integrin-recognition motif in active loop-1, did not adversely affect the tracer's ability to bind α v β 6 with highaffinity. The half-maximal inhibitory concentration (IC 50 ) values for the unlabeled precursor and FP-labeled standard were 0.61 ± 0.31 nM and 0.56 ± 0.46 nM S.D., respectively (Fig. 1c insert, Supplementary Fig. 3).
In order to select the most promising ligand for clinical translation,  Table 3). At 1 h post injection (p.i.), accumulation of tracer in the tumors were comparable at~3% injected dose per gram (%ID g −1 ) across the variants (Table 3). However, [ 64 Cu]DOTA-R 0 1-MR was eliminated from further consideration because its kidney uptake (~90 %ID g −1 ) was approximately 3-fold higher than the other two candidates (~30 % ID g −1 , Table 3 and Supplementary Table 2). Based on these initial studies, R 0 1-MG was selected as the lead translational candidate and evaluated as a radiofluorinated PET tracer.
[ 18 F]FP-R 0 1-MG-F2 was distributed broadly throughout the tumor (ROI~8000 mm 3 ) and resulted in a SUV mean of 6.2 at 1 h p.i. (Fig. 3a, Supplementary Movie 4). Comparatively, a region of focal [ 18 F]FDG uptake (ROI~3000 mm 3 , SUV mean = 4.1) was present adjacent to a biliary stent (Fig. 3b, Supplementary Movie 5). Tumor uptake is also shown in the volume rendered PET/CT images (Fig. 3a, Table 11). Sequential MIP PET images of a representative study subject acquired from 9 to 61 min after intravenous tracer administration (Fig. 4a)    . In one pancreatic cancer patient, a continuous band of tracer uptake occurred in a portion of the pancreas that spanned the head, uncinate process, neck, and tail (Fig. 4d). Here, the majority of the tumor mass occurred in the neck of the pancreas, but tracer uptake was seen in only a portion of that enlarged mass. The pathology report indicated moderately to poorly differentiated pancreatic adenocarcinoma, as well as a significant amount of necrosis in that tissue. H&E staining and IHC confirmed the expression of α v β 6 in the viable parts of the tumor (Fig. 4e).
In cervical cancer patients, relatively low tracer uptake in the lower large intestines (SUV mean = 1.54 ± 0.28) allowed for straightforward detection of cervical tumor where the average SUV mean = 4.79 ± 0.37 (Fig. 5, Supplementary Fig. 13  to similar levels (SUV mean = 4.1) in the tumor of the second cervical cancer patient ( Supplementary Fig. 13). H&E staining and IHC confirmed the expression of α v β 6 in the larger diffuse cancer cells shown in the tissue slice (Fig. 5b). Additionally, in one patient with lung cancer, very low uptake in the normal lung, SUV mean = 0.41 ± 0.11, led to high contrast of a lung tumor (Fig. 5c, SUV mean = 2.20) with a~5:1 ratio of lung tumor to normal lung.    (Fig. 6a, b, magenta arrows, reticulation, honeycombing, and basal/subpleural predominance). IPF-5 is a 72 y.o. male IPF patient who was diagnosed by CT with a definite-UIP pattern in 2015 (Fig. 6a). The highest levels of [ 18 F]FP-R 0 1-MG-F2 were found in the basal and subpleural regions of the lungs, consistent with the CT-UIP diagnosis as shown by the high degree of reticulation and honeycombing (magenta and cyan arrows). In contrast, the healthy portions of IPF-5's lungs were relatively devoid of the PET tracer (white arrows). The SUV mean s for the entire left and right lungs were 2.44 ± 1.16 and 1.90 ± 1.11, respectively (Fig. 7, Supplementary  Table 13). Due to the heterogenous distribution of fibrosis in ILDs, the extent of disease may be quantified by the SUV range (SUV max − SUV min , 6.82 ± 0.14 S.D.) as well as the fractional amount of damaged lung tissue ( Supplementary Fig. 14B). While the SUV max s for the left and right lungs were 6.93 and 7.07, the lung regions most severely affected by the disease are comprehensively described by the distribution of SUVs (SUV histogram) in the upper part of the SUV range (Fig. 6d). For the left lung,~18% of the SUVs were between 3 and 3.9, while~10% of the SUVs were >4 as shown in the PET tracers SUV histogram ( Supplementary Fig. 14B) and a summary of SUV metrics in the knottin PET tracer's fibrosis spectrum (Supplementary Table 13). For the right lung,~13% of the SUVs were between 3 and 3.9, while~5% were >4.
IPF-4 is a 72 y.o. male IPF patient who received a left lung transplant at Stanford Hospital in 2016. (Fig. 6b). Chest CT shows extensive fibrosis in the native right lung where [ 18 F]FP-R 0 1-MG-F2 accumulation was high and sustained (Fig. 6b). Accumulation of the PET tracer correlated with the CT-based UIP pattern of fibrosis in the right lung, where the SUV mean and SUV range were 1.89 ± 0.74 and 5.10, respectively. In contrast, the SUV mean and SUV range of the transplanted left lung were 0.93 ± 0.38 and 4.34, respectively (Fig. 7, Supplementary Fig. 14A, and Supplementary  Table 13). IPF-4's transplanted left lung is relatively devoid of the PET tracer, and the CT images are also consistent with a typical healthy lung (Fig. 6b).
In order to calibrate the IPF lungs to a reference standard, the lungs of the five healthy volunteers were similarly quantified. The within-subject intraclass correlation between lungs in healthy volunteers was 0.99 compared to 0.76 in IPF patients 34 . Due to  Table 13). The SUV-histograms of young lungs exhibited a Gaussian distribution (Fig. 6d) as modeled by baseline-20's normal distribution (Fig. 6d). In contrast, the highest [ 18 F]FP-R 0 1-MG-F2 levels (SUV mean = 1.41 ± 0.38) were found in HV-1, a 48-year-old currently asymptomatic male, who is a long-term bird-owner. Exposure to birds is a well-studied risk-factors for fibrotic ILDs 35,36 . The two remaining volunteers, both females in their 40s, showed mildly higher levels of [ 18 F]FP-R 0 1-MG-F2 relative to baseline-20 ( Fig. 6d, top, Supplementary Table 13).

Discussion
Our manuscript describes the complete process of developing and translating a cystine knot PET tracer for detection of multiple α v β 6 -positive indications, including cancer and IPF. In both preclinical small animal models and human clinical trial participants, accumulation of the knottin PET tracers with α v β 6 -positive tumors and the fibrotic portions of lungs was rapid, high, and sustained, compared to healthy tissues. Human cancer was detected in multiple regions of the body including the thorax (lung cancer), upper abdomen (pancreatic cancer), and lower abdomen (cervical cancer). Compared to very low background levels of this tracer in healthy young lungs, elevated tracer levels and a robust dynamic range of uptake were observed in the lungs of patients diagnosed with IPF. Accumulation of our PET tracers in most normal human tissues was generally low with notable exceptions in the upper gut and pituitary gland. Our initial findings suggest that the α v β 6 knottin PET tracers may find broad clinical application in the detection of diseases marked by overexpression of α v β 6 .
α v β 6 belongs to a subfamily of integrins called the RGD integrins 3,37 . Proteins or peptides that contain RGD motif are able to discriminate between different integrins by structural differences in the amino acids that flank the core RGD motif. The foot and mouth disease virus (FMDV) evolved an RGD flanking sequence in a viral coat protein that selectively recognizes α v β 6 with high-affinity, while molecularly ignoring the other integrins such as α v β 3 27 .
Cystine knot peptides engineered for selective, high-affinity molecular recognition of cancer markers have recently shown promise in pre-clinical models, not only as PET tracers, but in near infrared fluorescence imaging, molecularly targeted ultrasound imaging, and photoacoustic imaging 4,41,42 . Studies of knottin PET tracers have revealed that their pharmacokinetics are highly tunable through simple amino acid substitutions in their loops 26 . This pharmacokinetic versatility is one characteristic that distinguishes cystine knot PET tracers from other peptide-tracers such as the linear FMDV derivatives, simple cyclic peptides (cyclo-RGD) or peptides that depend on their framework amino acids (non-binding) for structural stability (triple helix bundle). Comparatively, these simple linear or cyclic peptides have shown improved pharmacokinetic profiles through chemical modification strategies such as dimerization and/or PEGylation 38,43 .
Three different imaging labels ([ 64 Cu]DOTA-, [ 68 Ga] NODAGA-, and [ 18 F]FP-) were used to evaluate the preliminary set of high-affinity knottins. Due to the relatively long half-life of 64 Cu (12.7 h), [ 64 Cu]DOTA-is well suited for initial screening of ligands beyond 2 h so that we may survey tracer uptake in tumor and normal tissues for up to 24 h after injection in pre-clinical models. In this case, once we determined that there was no benefit to imaging beyond a certain timepoint, shorter-lived radioisotopes with higher positron yields and lower overall radiation exposure were used. The equipment (e.g. cyclotron) and expertize at Stanford radiochemistry facilitates the development of [ 18 F]FP-R 0 1-MG-F2 and over a dozen clinical-grade 18 F labeled PET tracers for routine clinical use. However, the [ 68 Ga]NODAGA-R 0 1-MG version evaluated in this study by PUMC hospital may potentially be better suited for deployment due to the ubiquitous availability of 68 Ga (not requiring a dedicated cyclotron), and the ease of radiosynthesis through a pre-prepared kit. For these reasons, both the 18 F and 68 Ga versions were studied. R 0 1-MG PET tracers were evaluated in several different α v β 6positive cancers. In pancreatic cancer, accumulation of the PET tracer, [ 18 F]FP-R 0 1-MG-F2 was rapid and remained high throughout the study (SUV mean~6 at 1 h p.i.). Here, we were fortunate to have [ 18 F]FDG data from a clinical PET/CT for comparison. Accumulation of [ 18 F]FDG by the tumor was concentrated (~3000 mm 3 , SUV mean~4 at 1 h p.i.) near a biliary stent which was implanted to mitigate tumor-compression of the common bile duct. Comparatively in this same patient, the uptake of [ 18 F]FP-R 0 1-MG was more uniformly distributed over a greater tumor volume (~8000 mm 3 ). The difference in uptake profiles between the two classes of PET tracers probably reflects the different activities that are targeted (glucose metabolism vs. expression of ECM protein). In another pancreatic cancer patient, we observed a continuous band of intense uptake of [ 68 Ga]NODAGA-R 0 1-MG from the head of the pancreas (SUV mean~3 .8) to the tail (SUV mean~7 ), while the main tumor mass was located at the neck of the pancreas. Interestingly, the uptake did not occur throughout this entire tumor mass; tracer accumulation occurred only within the craniad of the enlarged mass as part of that continuous band of head-to-tail uptake. The pathology report indicated significant necrosis within the resected mass, and α v β 6-positive IHC staining occurred only in the viable part of the tumor. Pancreatic tumors in three patients were easily discernable from their PET/CT images despite high tracer accumulation in neighboring stomach and small intestine.
Accumulation of the knottin PET tracer was observed in the primary tumors of cervical cancer patients (SUV mean~5 ) where uptake by neighboring lower bowel and uterus were relatively  Table 13). The right/left correlation is 0.90 and 0.70 for HVs and IPF patients, respectively. The average SUV mean s ± S.D.s for the right lungs of HVs (n = 5) and the IPF group (n = 6) is 0.92 ± 0.33 and 2.04 ± 0.91, respectively (p = 0.0087). Source data are referred to in the Source Data File low. Due to the spatial resolution of the clinical PET scans, cervical cancers were clearly delineated from adjacent bladder containing urine-excreted knottin PET tracer. However, one caveat comes from an early report that found a 10-fold higher expression of β6 integrin subunit's mRNA in the stratum functionalis of endometrial epithelium 44 . Much lower mRNA levels were reported for the analogous specimen during the proliferative phase endometrium, suggesting that α v β 6 levels may be in flux during the menstrual cycle. Finally, very low α v β 6 knottin PET tracers accumulation in the upper thorax of normal subjects allowed the detection of lung cancer in one patient (Fig. 5c). This pilot clinical study with the knottin PET tracers for cancer detection is limited mainly by the small number of cancer patients. Therefore, recruitment of additional cancer patients is ongoing as we aim to comprehensively evaluate these knottin PET tracers for the cancers described here along with serous ovarian cancer, and head and neck cancer, both of which demonstrate elevated α v β 6 levels.
α v β 6 's role in TGF-β1 activation/signaling and in the pathogenesis of fibrotic diseases originating in lung and liver has been well documented [45][46][47][48] . Inhibition of fibrosis through molecular recognition of α v β 6 has been achieved in the bleomycin mouse model as well as in human clinical trials by a small molecule inhibitor and several monoclonal antibodies 28,49,50 .
[ 18 F]FP-R 0 1-MG-F2 enables whole lung molecular assessment of lung fibrosis through expression of α v β 6 , an ILD histological marker that correlates with prognosis 2 . The knottin PET tracer's fibrosis spectrum (Supplementary Table 13 provides information about the extent of disease, with IPF patients exhibiting rightward shift of the SUV mean , extension of the SUV range , and high-SUV voxels populating the upper bins of the SUV histogram . The ability to non-invasively characterize active fibrosis in ILDs including IPF, based on molecular expression may be useful for treatment planning and monitoring. A lower range of [ 18 F]FP-R 0 1-MG-F2-accumulation was observed in the lungs of healthy volunteers, which suggests that this PET-tracer may be able to discern healthy individuals from IPF patients 51,52 . Data for the low (healthy) end of the knottin PET tracer's fibrosis spectrum were obtained from the two youngest healthy volunteers (HV-3 and HV-5) in their twenties, who provided a hypothetical reference SUV baseline-20 mean and normal distribution for young lungs. Statistical departure (z-score) away from baseline-20 appears to correlate with the gradual transition from healthy lungs to fibrotic lungs in the case of IPF, as shown in the fibrosis spectrum ( Supplementary Fig. 12). The mildly increased levels of PET-tracer accumulation in the lungs of the two healthy mid-40s female volunteers (HV-2 and HV-4,~3σ from baseline-20 SUV mean ) suggests that α v β 6 levels increase with age. Interestingly, the high end of [ 18 F]FP-R 0 1-MG-F2's fibrosis spectrum from the healthy volunteers group (7σ from baseline-20 SUV mean ) was observed in a 48-year-old asymptomatic male (HV-1), who is a long-term bird-owner with a history of chronic second-hand-smoke and wood-dust exposure 35,51 . Although his low-dose chest CT showed relatively normal lungs, [ 18 F]FP-R 0 1-MG-F2 levels in HV-1 (SUV mean~1 .4) were higher and overlapped with IPF-3's SUV metrics in the lower range of the fibrosis spectrum (SUV mean~1 .1, Supplementary Table 13).
The possibility that [ 18 F]FP-R 0 1-MG-F2 can monitor dynamic changes in fibrosis was also suggested by patient IPF-4's successfully transplanted left lung, where tracer uptake levels were similar to levels recorded in lungs from healthy volunteers (Fig. 7). α v β 6 plays a key role in wound healing, and accumulation of the PET tracer is expected if the administration of [ 18 F]FP-R 0 1-MG-F2 had occurred during active tissue remodeling 53 . Interestingly, [ 18 F]FP-R 0 1-MG-F2 did not accumulate at the bronchial anastomosis site, suggesting the wound-healing process was complete 2-year post-transplant when the PET study was conducted.
The current pilot study for IPF was limited by the lack of histological data for patient-matched tissues. Although the expression of α v β 6 in IPF has been immunohistologically confirmed in multiple previous studies, patient-matched IHC studies may provide additional insight into correlation between PET tracer uptake and severity of disease 2,28 . Finally, our study does not contain enough clinical data to determine statistical correlation to other metrics used in IPF such as pulmonary function tests.
The accumulation of the knottin PET tracers occurred in the pituitary gland of all study subjects (where the PET scan included the brain, n = 11); this was an unexpected finding in our study. Although accumulation of our tracers did not occur in the other major regions of the normal human brain, we observed rapid and sustained uptake of our α v β 6 probes in the pituitary region throughout the duration of the study (SUV mean~4 at 1 h p.i.). A review of the literature indicated that several integrins are expressed in the normal pituitary gland, and that the expression pattern of integrins expressed in the pituitary changes with the occurrence of adenomas, which represent about 10-15% of all intracranial neoplasms [54][55][56] . Most of the studies that surveyed integrin expression in the pituitary were conducted many years earlier with a limited repertoire of antibodies for IHC, and before most investigators ever looked for the expression of the β6 integrin subunit in the pituitary gland. Therefore, there are several reports that provide a list of integrin subunits that were investigated at that time except for the β6 subunit. We were not able to find any current studies which confirmed, either at the message or protein level, the expression the β6 subunit in the pituitary. Moreover, there are several conflicting reports about the expression of the α v subunit in the healthy and diseased pituitary 54,57 . A parallel line of potentially related research has focused on the expression of TGF-β1 expression in the pituitary gland. Local TGF-β1 expression has been confirmed in both normal and cancerous pituitary tissues 55,58,59 . The reason that this is important for our studies is that expression of TGF-β1 is often associated with the expression of its activator, α v β 6 29,60 . Accumulation of our probe may have provided evidence for the expression of α v β 6 in the human pituitary. Although the anterior portion of pituitary gland remains outside of the blood brain barrier, our studies show that it is possible to target a 4 kDa peptide probe to the pituitary gland, which may allow the detection of adenomas or prolactinomas by developing probes that target cell surface proteins, such as other integrins that are specific for those diseases. Moreover, because of the sustained uptake of the knottin PET tracers in the pituitary, our studies suggest that it may be possible to exploit knottins, such as R 0 1-MG, for delivery of therapeutic activities to that region of the brain.
In conclusion, we have addressed several unmet clinical needs by developing cystine knot PET tracers that effectively detect multiple cancers in different regions of the body, as well as fibrotic changes in the lungs in IPF patients. In pancreatic, cervical and lung cancer, and in IPF, the knottin PET tracers demonstrated rapid and sustained accumulation in diseased tissues with relatively low background uptake in healthy organs or regions of the body prone to different cancers or fibrosis (lung and liver). These results suggest that the R 0 1-MG based PET tracers will have broad clinical application in detecting/diagnosing multiple indications, monitoring the efficacy of multiple therapeutics, as well as in staging both cancer and pulmonary fibrosis. The results from these pilot clinical studies encourage comprehensive evaluation of these α v β 6 cystine knot PET tracer across a broad range of disease states and applications in different patient populations.
Site directed evolution. The open reading frames encoding cystine knot peptides were generated by overlap-extension PCR using yeast-optimized codons defined by Lasergene (DNASTAR.com). The position that was randomized, as denoted by the letter x in Supplementary Fig. 1, was constructed with the NNB degenerate codon sequence. PCR products were amplified using primers with overlap to the pCT yeast display plasmid, which were upstream or downstream of the NheI and BamHI restriction sites, respectively. For each library,~40 µg of DNA insert and 4 µg of linearized pCT vector were electroporated into the S. cerevisiae strain EBY100 by homologous recombination. Electroporation was performed using cuvettes with a 2 mm gap. The electroporator was set to exponential decay mode, 540 mV and 25 µf.
The library was incubated and screened at room temperature for 2 h in 1 nM of recombinant integrin α v β 6 in IBB 31 . Next, a 1:250 dilution of chicken anti-cMyc IgY antibody (AB_2535826, Cat # A-21281, ThermoFisher) was added for 1 h at 4°C. The cells were washed with ice-cold IBB and incubated with a 1:25 dilution of fluorescein-conjugated anti-human α v integrin antibody (Clone NK1-M9, Cat # 327908, Biolegend) and a 1:100 dilution of Alexa 555-conjugated goat anti-chicken IgG secondary antibody (AB_2535858, Cat # A-21427, ThermoFisher) for 0.5 h at 4°C. Cells were washed in IBB and α v β 6 integrin binders were isolated using a Becton Dickinson FACS Aria III instrument. A diagonal sort gate was used to isolate yeast cells with enhanced integrin binding (FITC fluorescence) for a given protein expression level (Alexa 555 fluorescence). Plasmid DNA was recovered by Zymoprep (Zymo Research), amplified in Max Efficiency DH5α E. coli cells (Invitrogen) and sequenced.
Determination of equilibrium dissociation constants. Various concentrations (100 nM to 300 pM) of recombinant integrin α v β 6 were incubated with 10 5 yeast cells expressing R 0 1, R 0 1-MG, R 0 1-MR, or R 0 1-MW in the presence of 10 6 uninduced yeast cells. Prior to flow cytometry and analysis, yeast cells were processed, stained, and washed as described above in the section called Library Synthesis and Screening. Using two color flow cytometry, the binding of knottin (FITC) to integrin α v β 6 was normalized to expression level (Alexa 555) prior to determination of equilibrium dissociation constants (K D ). The normalized binding was plotted against the log of the concentration of recombinant integrin α v β 6 . The K D was determined by nonlinear regression analysis using Kaleidagraph (Synergy Software).
Determination of IC 50 s of knottin derivatives. Various concentrations of synthetic knottin peptide R 0 1-MG and its derivatives were incubated with 10 nM integrin α v β 6 at room temperature overnight. In order to determine the halfmaximal inhibitory concentration, the ligands, R 0 1-MG, DOTA-R 0 1-MG, NODOGA-R 0 1-MG, [ 19 F]FP-R 0 1-MG-F1 and [ 19 F]FP-R 0 1-MG-F2 were allowed to compete with 10 5 induced yeast cells (in the presence of 10 6 un-induced carrier yeast cells) surface-displaying R 0 1-MG for binding to recombinant integrin α v β 6 . A ten-fold molar excess of test ligand, relative to the moles of yeast surface displayed R 0 1-MG, was provided to the system (~50,000 R 0 1-MG/yeast cell).
Peptide synthesis and folding. Precursor peptide R 0 1-MG was synthesized on a CS Bio CS336 instrument via 9-fluorenylmethoxycarbonyl (Fmoc)-based solid phase peptide synthesis methods, and a Rink amide resin (CS Bio). Fmoc-protected amino acids were purchased from CS Bio. Fmoc groups were removed with 20% piperidine in DMF. Amino-acid coupling was performed using HOBt/diisopropylcarbodiimide (DIC) chemistry in DMF. Side-chain deprotection and resin cleavage was achieved by addition of a 94:2.5:2.5:1 (vv −1 ) mixture of trifluoroacetic acid (TFA)/trimethylsilane/ethanedithiol/water for 2 h at room temperature. The crude peptides were precipitated with cold, anhydrous diethyl ether and purified to >95% purity by semi-preparative reversed-phase HPLC using a Dionex Ultimate 3000 HPLC system and a Vydac 218TP1010 C18 column. Linear gradients of 90% acetonitrile in water containing 0.1% (v v −1 ) TFA were used for all peptide purifications, which were monitored at an absorbance of 220 nm. Peptide purity was analyzed by analytical reversed-phase HPLC using either a 214TP C4 5μ or Phenomenex Aeris C 18 column. Molecular masses were confirmed by matrix-assisted laser desorption/ionization-mass spectrometry (MALDI-MS; ABI 5800, Supplementary Table 1).
Precursor peptide folding reactions were performed by gently rocking peptides for 12-20 h in a 0.1 M ammonium bicarbonate, pH9 solution containing 2.5 mM reduced glutathione, 20% dimethlysulfoxide (v v −1 ), 20% isopropanol (vv −1 ), and 20% 0.8 M guanidinium hydrochloride. The final oxidized (folded) precursor product was purified by semi-preparative reversed-phase HPLC as described above. Following purification, folded peptides were lyophilized and stored at room temperature until used. Purified peptides were dissolved in water, and concentrations were determined by amino acid analysis (Jay Gambee AAA Service Laboratory, Damascus, OR). Peptide purity and molecular mass were confirmed by analytical reversed-phase HPLC and MADLI-MS. The CGMP precursor peptide R 0 1-MG was made, using a similar protocol, by CS Bio (Menlo Park CA).
DOTA or NODAGA conjugation. Briefly, 1-2 mg of peptides were conjugated tõ 2 mg DOTA-NHS (Macrocyclics, Plano TX) or NODAGA-NHS (Chematech, Dijon France) in 500 μL DMF containing 2 μL DIPEA at room temperature for up to 1 h. Chelator-peptide conjugates were acidified in solvent A purified by semipreparative reverse phase HPLC as described above. 64 Cu radiolabeling. Approximately 10 μg of peptide were combined with~2 mCi 64 CuCl 2 in~500 μL 100 mM sodium acetate buffer pH5.5 at 37°C with gentle shaking at 250 rpm for at least 1 h prior to purification by PD-10 column. Twelve 500 μL fractions were collected from the column. The active fraction (typically 6 and 7) was subsequently used for in vivo studies. 68 Ga radiolabeling. 68  Tumor models. We have complied with all relevant ethical regulations for animal testing and research. Animal procedures were performed with approval from the Stanford University Administrative Panels on Laboratory Animal Care (APLAC, protocol #9536). Female athymic nude mice, 4-6 weeks old (Charles River), were subcutaneously shoulder-injected with 10 7 cells suspended in 100 µL PBS. Mice were used for imaging and biodistribution studies when xenografted tumors grew to a diameter of~1 cm.
Static small animal PET imaging. Tumor-bearing Nu/Nu female mice (n = 3 for each probe) were anesthetized using 2% isoflurane in oxygen and injected with 100 µCi (~0.15 nmol) of the tracers described above via the tail vein. Five-minute static PET scans were acquired on an Inveon PET-CT or Inveon D-PET scanner (Siemens Healthcare, Malvern PA). Images were reconstructed by a twodimensional ordered expectation maximum subset algorithm and calibrated as described below. ROIs were drawn over the tumor on decay-corrected whole-body images using Inveon Research Workplace (IRW) software (Siemens) or ASIPro VM software (Siemens). ROIs were converted to counts g −1 min −1 , and %ID g −1 values were determined assuming a tissue density of 1 g ml −1 .
Calibration of small animal PET. Scanner activity calibration was performed to map between PET image units and units of radioactivity concentration. A preweighed 50-mL centrifuge tube was filled with distilled water and 64 CuCl 2 (∼9.3 MBq [∼250 μCi] as determined by the dose calibrator) was used to simulate the whole body of the mouse. This tube was weighed, centered in the scanner aperture, and imaged for a 30-min static image, single bed position. From the sample weight and assuming a density of 1 g ml −1 , the activity concentration in the bottle was calculated in units of μCi mL −1 . Eight planes were acquired in the coronal section. A rectangular region of interest (ROI) (counts/pixel/s) was drawn on the middle of eight coronal planes. Using these data, a calibration factor (C) was obtained by dividing the known radioactivity in the cylinder (μCi mL −1 ) by the image ROI. This calibration factor was determined periodically and did not vary significantly with time.
Dynamic small animal PET and PET/CT imaging. Dynamic scans were acquired over~2 h p.i. Image acquisition was initiated 15 seconds prior to tracer injection and proceeded for 115.1 min p.i. The acquired data were then sorted into 0.5-mm sinogram bins and 26 time frames for image reconstruction (4 × 15 s, 1 × 37.5 s, 4 × 60 s, 11 × 300 s and 5 × 600 s), which was done by iterative reconstruction using the following parameters: 3D ordered-subsets expectation maximization (3D-OSEM) followed by fast maximum a posteriori (fastMAP); MAP OSEM iterations: 2; MAP subsets: 16; MAP iterations: 18. ROI analysis (IRW) was performed on the tumor and the major organs (heart, liver, kidneys, bladder, and muscle) seen on the dynamic PET scan images. The count densities were averaged for all volumes of interest at each timepoint to obtain a time versus radioactivity curve (TAC). Tumor TACs were normalized to injected dose, measured by a CRC-15 PET dose calibrator (Capintec, Inc.), and expressed as percentage injected dose per gram of tissue (%ID g −1 ), assuming 1 g ml −1 .
Pre-clinical radiation dosimetry. Non-decay-corrected uptake (%ID g −1 ) values from the dynamic PET study (above) were converted to %ID organ −1 and then subjected to an animal-to-human biokinetic extrapolation using the percent kg g −1 method where (%ID organ)human = [(%ID g −1 ) animal × (kg TBweight ) animal × g organ (kg TBweight ) human −1 ]. The animal whole body weight was 34 g and the weights of the human organs were derived from a 73 kg male and 58 kg female in Organ level Internal Dose Assessment (OLINDA) software; source organ residence times were calculated using a bi-exponential model OLINDA 61 . The projected human doses were then computed for human phantoms using the source organ residence times.
In vitro and in vivo stability. Aliquots of [ 18 F]FP-R 0 1-MG-F2 were incubated in an equal volume of AB Human Serum (Invitrogen, 34005100) for 2 h at 37°C. Samples were acidified with TFA and centrifuged at 18,000 g for 3 min to remove precipitants. For in vivo stability, mouse urine was collected with a syringe immediately after euthanasia. Typically,~100 μL of urine was released by the bladder as it relaxed when the mice died; urine pooled at the genitals and was stabilized by surface tension. All samples were analyzed by radio-HPLC on a Dionex C 4 column. The following describes part 1 of the [ 19 F]FP-R 0 1-MG-F2 Synthesis, the synthesis of the [ 19 F] Nitrophenyl Ester (NPE). 2-Fluoropropionic Acid (FPA) was mixed with 1 M NaOH in a 1:1 ratio and allowed to react at 95°C for 10 min. The reaction mixture was transferred to a round bottom flask and two volumes of acetonitrile was added to the mixture for azeotropic distillation at 30°C on a Rotary Evaporator. The sodium 2-fluoropropionate (FP -Na + ) product appeared to be a white flaky powder with a molecular weight of 130.15 g n −1 . Its mass and structure were verified by ESI-MS and NMR, respectively.
Next, sodium 2-fluoropropionate (solid) was mixed with 328 mM bis(4nitrophenyl) carbonate (NPC) in acetonitrile in a 1:2 to 1:7 molar ratio. The round bottom flask was placed in 90-110°C oil bath for 20-30 min. The resulting product, 2-fluoro nitrophenyl propanoate referred to as the nitrophenyl ester ([ 19 F] NPE) compound was purified by semi-preparative RP-HPLC using a Phenomenex Gemini C18 column. A 90% acetonitrile, 0.1% trifluoroacetic acid (TFA, solvent B) gradient was used. The solution containing purified [ 19 F]NPE was next diluted in a 1:1 (vv −1 ) ratio with solvent A (99.9% H2O, 0.1% TFA) and loaded onto a Waters Sepak C18 column, which was rinsed three times with solvent A. The purified product, [ 19 F]NPE was eluted from the Sepak into a 20 mL glass storage bottle using 2 mL diethyl ether. The ether was evaporated on a hot plate (80-90%) in the presence of an air stream, which resulted in dry [ 19 F]NPE as a white crystal film.
X-ray data collection of [ 19 FN]FP-R 0 1-MG-F2. X-ray diffraction experiments were performed at the MBC Beamline 4.2.2 of the Advanced Light Source using the RDI-8 m CMOS detector 74 . Crystals were tested for diffraction using a Superbend magnet source coupled to a Rosenbaum-Rock Si(111) sagitally focused monochromator with an energy range of 5500-16,000 eV. An ACTOR robot (Rigaku) was used to load frozen crystals into position on the beamline. A number of CSHToptimized D4 crystals (50-200 μm) were screened and produced high-quality diffraction data beyond 1.0 Å resolution. For each crystal, 180 degrees of data were collected in shutterless mode with 0.1 degree frames at an energy of 1.00 Å and a temperature of 100 K; if necessary, a second dataset was collected on the same crystal at 50% attenuation to record overloaded reflections from the first pass. Datasets from individual crystals were indexed, integrated, scaled, and merged in X-ray structure determination of [ 19 F]FP-R 0 1-MG-F2. In order to solve the structure, we used a high-quality dataset obtained from a crystal that diffracted to 1.05 Å ( Table 2). The phases for [ 19 F]FP-R 0 1-MG-F2 were solved using a molecular replacement search model made from the NMR structure of the parent peptide scaffold Momordica Cochinchinensis Trypsin Inhibitor II (MCoTI-II, 2IT8.pdb and 2N8B.pdb) subsequently engineered by our group to develop the precursor peptide R 0 1-MG. MCoTI-II and R 0 1-MG share~60% sequence identity; they are different in their active Loop-1 ( Supplementary Fig. 2). The initial search model was generated by deletion of the active 2000-1 and conversion of all other non-identical (non-loop) residues to serine. Molecular replacement program PHASER, which is a part of the CCP4 crystallography program suite, was used to search two molecules in the asymmetric unit (ASU) 76,78 . An initial solution with the search model was obtained in P2 1 2 1 2 1 space group.
The model-building program ARP/wARP was used to automatically construct more than 90% of the model in the ASU 79 . A complete model was obtained by further cycles of manual refinement and loop building in Refmac5 (CCP4 suite) and COOT 80 . Once all the amino-acid residues were fitted to the electron density map, the N-terminus fluropropyl group was manually determined by using the ligand generating tools in the CCP4 suite and COOT. Further refinement cycles led to a final molecule, 6CDX.pdb, with an R-Factor (R work ) of 0.18 and a resolution of 1.05 Å. Healthy volunteer inclusion criteria. Volunteers met all of the following inclusion criteria and were considered eligible for participation in this study.   Healthy volunteer exclusion criteria. If volunteers met any of the following exclusion criteria, they were considered ineligible for participation in this study.
Less than 18-year old at the time of radiotracer administration Pregnant or nursing Pancreatic cancer patient exclusion criteria. If cancer patients met any of the following exclusion criteria, they were considered ineligible for participation in this study.
Less than 18-year old at the time of radiotracer administration Pregnant or nursing Contraindications for PET/CT Unable complete a PET/CT scan Unable to comply with the study procedures Serious uncontrolled concurrent medical illness that would limit compliance with study requirements Eastern Cooperative Oncology Group Performance Status (ECOG PS) > 2 IPF patient exclusion criteria.
Patients with serious uncontrolled concurrent medical illness that would limit compliance with study requirements. Patient has history of any clinically significant lung disease other than IPF as determined by pulmonologist. Patient has had a lung infection of any kind in the last 3 months.
1. Participant was asked to drink 1-2 glasses oher arrival at the clini/her arrival at the clinic. 2. Participants was consented by the responsible physician. 3. Female participants had an early pregnancy test (EPT) to rule out pregnancy. 4. Participants was asked to urinate prior to start of study and instructed to collect a urine sample. 5. Participant was weighed and measured in height. 6. Participant had an IV line placed in an arm vein for tracer administration. 7. Participant had a second IV line placed in an arm vein for blood sampling. 8. A blood sample was taken 5 min prior to the injection of the tracer and used for baseline chemistry and hematology laboratory testing. 9. [ 18 F]FP-R 0 1-MG-F2 was formulated in a sterile and pyrogen-free isotonic solution and was administered in a single slow IV injection. The line was flushed with at least 10 ml normal saline after injection. 10. Blood samples for blood time activity measurements were taken at 1, 3, 5, and 10 min after tracer administration and then at 30-minute intervals for up to 3 h after tracer injection. The counts in whole blood and plasma were measured by gamma counter (Perkin Elmer Wizard 1470). All blood counts were decay corrected to the time of tracer injection. 11. A small low activity calibration source of known activity of 18  14. Static total-body (vertex-to-toe) PET/CT scans were acquired at 60 min and 120 min post injection. CT was performed at 120 kV; it was dose-modulated based on body habitus with current ranging from 10-105 mA for the 60minute scan. Attenuation correction was also used with CT for the 120minute scan (120 kV, 10 mA). 15. Blood pressure, temperature, heart rate and pulse oximetry measurements were taken before injection (baseline) and at 5, 10, 15, 60, and 120 min after  Table 10).
[ 18 F]FP-R 0 1-MG-F2 in IPF lungs. One female and five male study subjects, 72.5 ± 1 y.o., who were diagnosed with IPF according to international consensus diagnostic criteria of either a chest CT demonstrating a definite UIP pattern or lung tissue biopsy with a histopathologic UIP pattern, were administered 5-10 mCi of [ 18 F]FP-R 0 1-MG-F2 1 . One male study subject experienced dyspnea, elevated heart rate, coughing and shivering during the study. The female study subject reported shortness of breath and nausea after 48 h. Upon review by our nuclear medicine physicians and pulmonary physicians, it was determined that these potentially adverse reactions were not likely study-related, but rather a result of the underlying disease.
Analysis of radiotracer in study subject's blood. Blood samples for time-activity measurements were obtained at 1, 3, 5, and 10 min after tracer injection and then at 30-min intervals for up to 3 h after tracer injection. For each timepoint, 500 μL of blood was transferred to a 13 × 75 mm plastic tube containing 75 USP sodium heparin (Benton Dickinson Vacutainer 367871). A separate fraction of 500 μL aliquot of blood was spun in a clinical centrifuge for 5 min at room temperature. Two hundred microliter of the plasma fraction from this second aliquot was transferred to a new Vacutainer tube. In order to match the volume of the whole blood sample, water was added to the plasma fraction and to the cellular fraction to a final volume of 500 μL. The radioactive counts in whole blood (first fraction), and the plasma fractions were measured with a gamma counter (Perkin Elmer Wizard 1470). All blood counts were decay corrected to the time of tracer injection.
Dosimetry (OLINDA). Organ doses values were calculated using organ-level internal dose assessment (OLINDA) software (Vanderbilt University, 2003) 61 . The organs with highest uptake were included in the assessment. These include the kidneys, stomach, small intestine and bladder. A threshold-based segmentation on the last frame of dynamic PET images were used to determine ROIs for the kidney and bladder due to their high tracer uptake relative to background. PET/CT overlay images were used to manually trace ROIs of the stomach and small intestine. Organs with lower uptake, were also included in the dosimetry assessment including the pancreas, liver, lung, and heart. ROIs for these organs were also determined manually on PET/CT overlay images. Muscle uptake was also included in this study due to its large mass representation. In addition, a whole-body ROI was determined by threshold-based segmentation with the baseline threshold value set to exclude image noise occurring outside the body. ROIs were drawn in consensus (by R.H.K, L.B, and F.H.). Normalized cumulative activity was determined for each organ from the area under the curve (AUC) which was calculated using the trapezoidal rule and assuming a physical decay only after the last measurement divided by the total injected activity. These data were used to estimate the radiation doses absorbed by all of the organs.
Immunohistochemistry. Formalin-fixed paraffin-embedded tumor tissues were sectioned at 5 µm thickness. All histologic sections were stained with standard hematoxylin-eosin immunohistochemical staining (H&E). To confirm the presence of α v β6, additional sections were stained with immunohistochemistry (IHC) for anti-human α v β6 (0.625 μg ml −1 , 6.2A1, Biogen Idec, Cambridge, MA). For preparation of IHC staining, the slides were deparaffinized with xylene and rehydrated in serially diluted ethanol solutions (100-50%), followed by demineralized water according to standard protocols. Endogenous peroxidase activity was blocked by incubation in 0.3% hydrogen peroxide in phosphate buffered saline (PBS) for 20 min. Antigen retrieval for α v β 6 was performed with 0.4% pepsin incubation (Dako) at 37°C for 10 min. Following antigen retrieval, the tissue sections were incubated overnight with the primary antibodies in 100 µl 1% BSA in PBS at room temperature. The slides were washed with PBS, followed by incubation with secondary antibody at room temperature according to the Vectastain ABC HRP kit (Peroxidase, Mouse IgG, Vector Laboratories Inc., Burlingame, CA). After additional washing, the staining was visualized with 3,3-diaminobenzidine tetrahydrochloride solution (DAKO, Glustrup, Denmark) at room temperature for 5 min and counterstained with hematoxylin for 20 s. Finally, the tissue sections were dehydrated and mounted in Pertex (Histolab, Rockville, MD, USA). One patient reported nausea and vertigo~1 h p.i.; the complaints lasted about an hour and disappeared without treatment. The remaining six study subjects did not report any adverse reactions, and none were observed by the attending physicians and clinical researchers. Study participants did not prepare in any way (i.e. fasting) for the imaging study. Each patient, except for one (described below), underwent a single wholebody static PET/CT scan at 29-60 min after intravenous injection of [ 68 Ga] NODAGA-R 0 1-MG. The administered dose was 55.5-96.2MBq (1.5-2.6 mCi). The scan (6 bed positions, 2 min bed −1 ) covered an area that included the top of the skull to the middle of the femur. OSEM was applied for reconstruction. One patient with cervical cancer underwent multiple sequential whole-body static PET scans. For this cervical cancer patient, a single whole-body low-dose CT scan was performed prior to tracer administration, which was used for reconstruction and as an anatomical reference for the sequential PET images. The CT scan (140 kV, 35 mA, pitch of 1:1, layer of 5 mm thickness; layer spacing of 3 mm, 512*512 matrix, field of view of 70 cm) also covered the top of the skull to the middle of the femur. After the CT scan, the first whole-body PET scan (6 bed positions, 2 min bed −1 ) was acquired concurrent with the intraveneous administration of 77.7MBq (2.1 mCi) [ 68 Ga]NODAGA-R 0 1-MG. Immediately upon completion of the first scan, the second whole-body static PET data was acquired. Sequentially, multiple whole-body static PET scans were acquired, one by one, from 0 to 75 min. A total of 6 whole-body static PET scans (6 beds of each timepoint) were performed in this patient.
[ 18 F]FDG PET/CT (PUMC Hospital). All patients were instructed to avoid strenuous work or exercise for at least 24 h prior to the scheduled study date. Study participants were also instructed to fast for at least 4 h before intravenous administration of the PET tracer. Patients were administered [ 18 F]FDG at a dosage of 5.55 MBq (0.15 mCi) per kilogram body weight. Immediately after tracer administration, patients were allowed to rest and relax in a warm, darkened room for approximately 45-60 min. Next, each patient emptied their bladder. The PET scan that followed spanned a region from the mid-thigh area to the base of the skull (six bed positions, 2 min bed −1 ).

Region of interest.
Upon co-registration of the PET and CT images, the CT images were used to determine regions of interest for the organs or tissues reported in the manuscript. These ROIs were applied to the PET images of the brain, pituitary gland, breast, lower and upper large intestine, small intestine, stomach, pancreas, heart wall, liver, lung, muscle, red marrow, bone, skin, and thyroid gland. For high uptake organs such as the kidneys, bladder and for the whole body, threshold-based segmentation of PET images were used to determine a close-fitting ROI around the entire organ or whole body. In order to determine accuracy, these PET-derived ROIs were compared to the CT overlay for a close match. For the pancreatic lesions, several ROIs for the knottin vs. FDG tracers were compared. These data are reported as mean SUVs.
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-11863-w ARTICLE Statistical analysis. All data are presented as the average value ± the SD of at least three independent measurements. Statistical analysis for animal studies and binding studies were performed by two factor ANOVA without replication analysis using Microsoft Excel. Significance was assigned for p values of <0.05. Difference in SUV between healthy volunteers and IPF patients was tested by an exact Wilcoxon ranksum test using Stata Release 15.1 (StataCorp LP, College Station, TX). The correlation between left and right lungs was estimated by an average absolute-agreement intraclass correlation from a one-way random-effects model 34 . Due to high degree of correlation between the two halves of the lungs, only the right ones were used to avoid the transplanted left lung of IPF-4.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Source data for the following Figures, Supplementary Figs.  Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/.