PET imaging of fibroblast activation protein alpha (FAP) detects incipient cardiotoxicity due to anthracycline chemotherapy

ABSTRACT

Background Anthracycline chemotherapy is associated with a risk of cardiotoxicity leading to heart disease, particularly in pediatric cancer patients. Gold standard methods of detecting cardiotoxicity are insufficiently sensitive to early damage and specific pathophysiologies driving disease. Positron emission tomography (PET) couples anatomical resolution with biochemical mechanistic selectivity and potentially addresses the current diagnostic limitations in cardio-oncology. We aimed to validate PET imaging biomarkers targeting fibroblast activation protein alpha (FAP), Translocator protein (TSPO), and norepinephrine receptor (NET) for detection of incipient anthracycline-induced cardiotoxicity.

Methods Cardiotoxicity was established in male C57BL/6J mice by a cumulative dose of 24 mg/kg doxorubicin (DOX) over 2 weeks. DOX mice and their age-matched controls were imaged with echocardiography and PET, using [⁶⁸Ga]Ga-FAPI-04, [¹⁸F]DPA-714, and [¹⁸F]MFBG, over 12 weeks. Fractional shortening (FS) was determined from the echocardiograms, and cardiac uptake of the radioligands was quantified from the PET images. Heart sections were collected and used for the analysis of bulk RNA-seq, RT-qPCR, Western blot, in situ hybridization (ISH), and histopathological analysis.

Results DOX mice exhibited cardiotoxicity and cardiac atrophy. Cardiac [⁶⁸Ga]Ga-FAPI-04 PET signal was significantly higher in DOX mice from 2 weeks through the study endpoint. By contrast, no cardiac dysfunction was evident by echocardiography until 10 weeks, at which point FS was significantly reduced in DOX mice. There were no differences in [¹⁸F]DPA-714 and [¹⁸F]MFBG signals. Transcription and translation of FAP, but not TSPO or NET, was detected in cardiomyocytes and were elevated in the DOX hearts, in agreement with the PET data. Genes related to cell adhesion and extracellular remodeling were significantly upregulated in the DOX mice relative to controls.

Conclusions FAP is a sensitive and selective imaging biomarker for incipient cardiotoxicity and FAPI PET is a promising non-invasive imaging tool for identifying patients at risk of cardiotoxicity during or after anthracycline chemotherapy.

What is Known?

• Anthracycline chemotherapy results in cardiotoxicity for a sizeable population of treated patients. Cardiotoxocity manifests as cardiac dysfunction, and may result in long-term cardiac disease and heart failure, particularly in survivors of pediatric cancer.

• Cardiotoxicity is typically defined in terms of left ventricular ejection fraction (LVEF) deficits, as measured by echocardiography. However, this metric is often poorly sensitive to early disease and agnostic to underlying pathophysiology.

• Early treatment of cardiotoxicity improves recovery and long-term survival, emphasizing the need for accurate diagnostics in incipient disease.

What New Information Does This Article Contain?

• [⁶⁸Ga]Ga-FAPI-04 accumulates in the hearts of mice experiencing doxorubicin-induced cardiotoxicity as a function of fibroblast activation protein alpha (FAP) expression and activity. By contrast, cardiac uptake of radioligands targeting the translocator protein 18-kDa (TSPO) and the norepinephrine transporter (NET) do not differ between DOX animals and controls.

• Positron emission tomography (PET) imaging following administration of [⁶⁸Ga]Ga-FAPI-04 detects abnormal cardiac remodeling significantly earlier than LVEF decrease is observed, indicating that it may be more sensitive to incipient disease.

What New Information Does This Article Contain?Our study identifies fibroblast activation protein alpha (FAP) as a promising diagnostic imaging biomarker in anthracycline-induced cardiotoxicity. We show that cardiac PET signal increases immediately after doxorubicin treatment, and the signal increase is sustained for at least 10 weeks. In addition, we demonstrate that FAP inhibitor (FAPI) PET correlates with expression of FAP protein and gene. Thus, we provide mechanistic insight into potentially-treatable pathophysiologies driving cardiac atrophy and toxicity, and have identified a translational PET tracer that can image the activation of these processes at an early stage.

INTRODUCTION

Cancer therapy with the anthracycline doxorubicin (DOX) is the treatment of choice for a broad range of cancers, especially for the treatment of solid tumors and leukemias in adults and children.^1,2 Roughly 60% of pediatric cancer patients today receive DOX as part of their treatment.³ However, despite being a mainstay of anti-cancer therapy, DOX can induce cardiovascular dysfunction, especially cardiotoxicity that results in heart failure.^4,5 Heart conditions ranging from cardiomyopathy to heart failure are major adverse events in cancer patients treated with DOX, with up to 70% of total adverse events relating to cardiac health.^6,7 Childhood cancer survivors are particularly susceptible, with more than 10% of those who received DOX treatment developing cardiotoxicity,⁸ which can develop into severe heart disease in adulthood.⁹ Early diagnosis of heart failure can lead to less invasive and more effective treatment. Therefore, the identification of emerging cardiotoxicity and attendant cardiac damage is an essential need to improve outcomes in cancer treatment.

Cumulative DOX exposure can produce the symptoms of cardiotoxicity within weeks or months (“early” cardiotoxicity) or after a number of years (“late”).¹⁰ The specific cellular and molecular changes responsible for these responses have not been fully elucidated.¹¹ Assessment of left ventricular ejection fraction (LVEF) by echocardiography (echo) is currently the gold standard for evaluating cardiac function in patients with suspected cardiotoxicity.¹² However, echo is subject to temporal variability^13–15 and is poorly sensitive to early myocardial damage,^15,16 while measurement of LVEF alone discounts other cardiopathologic effects that may occur.¹⁷ The introduction of tissue Doppler and strain imaging echo has allowed subclinical cardiotoxicity to be detected more reliably through speckle tracking-based deformation analysis.^10,18 Nevertheless, the possibility of intervendor variability in strain measurements may require all follow up scans to be conducted using identical instrumentation.¹⁸ Furthermore, while improvements in echo image analysis have increased detection of subclinical cardiac functional decline, this method cannot be used to detect the underlying pathology responsible for these deficits. Cardiac magnetic resonance is an alternative imaging modality for assessing cardiac dysfunction,¹⁹ but this method still exhibits limited cost-effectiveness and uncertain diagnostic or prognostic value.^20,21 Nuclear imaging methods, such as single photon computed tomography (SPECT), and positron emission tomography (PET), have been developed for assessing LVEF, myocardial viability, and perfusion²² but these do not provide mechanistic insight into cardiotoxicity. Circulating cardiac troponins and B-type natriuretic peptides are blood biomarkers of cardiac injury that can be assessed through minimally invasive procedures,^21,23 however, their sensitivity and specificity may be insufficient to support use of these indices as single predictive biomarkers without further validation by an imaging technique,⁵ paticularly when samples are taken during early (acute) timepoints.²⁴ To identify at-risk patients prior to anthracycline chemotherapy and detect incipient cardiotoxicity arising during or after chemotherapy therefore requires the validation of new biomarkers related to specific disease-causing pathophysiologies.

PET is a promising modality with which to pursue this aim. PET imaging can be combined with other modalities such as computed tomography (CT) to allow functional and anatomical imaging with spatial resolution as high as 2 mm.²⁵ Recently, a number of radiolabeled small molecules for imaging fibroblast activation protein alpha (FAP),^26–29 translocator protein 18-kDa (TSPO),^30–32 and the norepinephrine transporter (NET)^33,34 by PET have been described for detecting cancer-associated fibrosis, inflammation, and sympathetic innervation, respectively. Among these PET agents, FAP inhibitor [⁶⁸Ga]Ga-FAPI-04,²⁷ TSPO ligand [¹⁸F]DPA-714,³⁵ and NET ligand meta-[¹⁸F]fluorobenzylguanidine, [¹⁸F]MFBG,³³ have undergone preliminary clinical evaluation for oncologic and neurologic applications. Cardiac tissue remodeling,³⁶ inflammation and cardiomyocyte mitochondrial dysfunction,³⁷ and loss of cardiac sympathetic innervation³⁸ are also identified as contributing pathologies in anthracycline-induced cardiotoxicity. Despite promising detection sensitivity, whole-body dosimetry, and relevance to key pathophysiologies of heart failure, neither these probes nor their molecular targets have been systematically evaluated in the context of cardiotoxicity.

Here, we established a preclinical model of DOX-induced cardiotoxicity for validating [⁶⁸Ga]Ga-FAPI-04, [¹⁸F]DPA-714, and [¹⁸F]MFBG as diagnostic and prognostic biomarkers in cardio-oncology. Our goal was to evaluate the ability of these radioligands to detection incipient cardiotoxicity and identify the most suitable probe for translational to clinical imaging of cancer survivors treated with anthracyclines.

METHODS

General

Doxorubicin hydrochloride was purchased from Tocris Bioscience, USA and used without further purification. It was dissolved at a concentration of 0.75 mg/mL in sterile saline for injection (Hospira, USA) with the aid of sonication. The solution was stored in the dark at -20 °C for up to 24 h before use.

Mouse model of doxorubicin-induced cardiotoxicity

All animal studies were approved by the Institutional Animal Care and Use Committee of Weill Cornell Medicine and were undertaken in accordance with the guidelines set forth by the U.S. Public Health Service Policy on Human Care and Use of Laboratory Animals. Adult male C57BL/6J (8-week-old) mice were purchased from The Jackson Laboratory, USA and randomly assigned to treatment (n=40) or control (n=16) groups. Mice in the treatment group were administered a solution of doxorubicin in saline at 3 mg/kg every other day for 2 weeks (total 8 doses; cumulative dose of 24 mg/kg) by intraperitoneal (i.p.) administration.³⁹ Age-matched control (CTRL) mice were administered the same volume of saline i.p. The mice were weighed three times weekly and given access to food and water ad libitum.

Echocardiography imaging and analysis

For evaluation of cardiac dimensions and function, echocardiography was performed under inhaled isoflurane anesthesia on a 37 °C heated platform using a Vevo 770 and 3100 Imaging systems (VisualSonics, Canada) accordingly to previously published methods.⁴⁰ Briefly, scans were acquired using left-ventricle M-mode and all measurements were obtained by averaging the values of three consecutive cardiac cycles. Left-ventricle end-diastolic (LVDd) and end-systolic (LVDs) dimensions were measured using M-mode traces. Fractional shortening (FS) was calculated using the formula [(LVDd-LVDs)/LVDd]. Diastolic measurements were estimated at the point of maximum cavity dimension, and systolic were taken at the point of minimum cavity dimension, according to the American Society of Echocardiography’s recommended method.⁴¹

Radiochemistry

[⁶⁸Ga]FAPI-04,²⁷ [¹⁸F]DPA-714,³⁰ and [¹⁸F]MFBG³⁴ were synthesized from their corresponding precursors according to literature procedures with minor modifications. Full experimental details describing the synthesis of the precursors and radioligands can be found in the Supplemental Material. All radioligands were formulated in 10% v/v ethanol/saline for administration to mice.

Small animal microPET/CT imaging

DOX and CTRL mice were intravenously administered 100-150 µL of a 10% v/v ethanol/saline solution containing 3.7-11.1 MBq of the corresponding radioligand. The mice were imaged in groups of 2-4 using small-animal microPET/CT (Siemens Inveon™, USA) under isoflurane anesthesia (3.5% for induction, 1.5 % for maintenance) beginning 45 min after injection. The total PET acquisition time was 30 min, and a CT scan was obtained immediately before the PET acquisition for anatomic coregistration and attenuation correction. Images were reconstructed using the commercial Inveon software provided by the vendor. Images were corrected for decay and for the total activity injected.

MicroPET/CT imaging data analysis

All microPET/CT images were evaluated with the AMIDE algorithm (A Medical Image Data Examiner).⁴² An ellipsoidal volume of interest (VOI) was generated for the heart and the right thigh muscle. The mean counts in the VOI were converted to percent injected dose per cubic centimeter (%ID/cm³) using the AMIDE algorithm, which was calibrated against a 1% injected dose standard. The %ID/cm³ in the heart was normalized against the %ID/cm³ in the muscle, providing a heart/muscle ratio, H/M.

Preparation of heart tissue

The mice were anesthetized by i.p. ketamine injection and perfused with phosphate-buffered saline (PBS) via the left ventricle at a constant pressure of 80 mmHg. The hearts were patted dry and weighed on a digital balance. To perform the molecular and histological analysis, the hearts were cut transversally at the mid-horizontal plane. Cut fractions were separated for RNA and protein extraction, and tissue staining. Regions for tissue staining were fixed overnight in formalin at 4 °C and stored in 70% ethanol until further processing. The fractions for RNA and protein extraction were flash frozen in liquid nitrogen and stored at -78 °C until further use.

Heart weight to tibia length (HW/TL) ratios

The tibia was collected ex vivo and all soft tissue was removed. The length was measured using a digital calipers. Heart weights were determined as described above and divided by the corresponding tibia length from the same animal to produce the heart weight to tibia length (HW/TL) ratio.

Western blotting

Frozen heart tissue was pulverized by mortar and pestle for tissue homogenization. Ground heart tissue powders were collected and soaked in tissue protein extraction reagent (#78510, ThermoFisher, USA) supplemented with a protease inhibitor cocktail (#87786, ThermoFisher, USA) for protein extraction. Protein concentrations and western blot (WB) were performed as previously reported.⁴³ The primary antibodies were anti-topoisomerase 2β) (TOP2β, 1:1000 dilution, #20549-1-AP, Proteintech, USA), anti-fibroblast activation protein alpha (FAP, 1:500 dilution, ab53066, Abcam, UK), anti-PBR (TSPO, 1:6000 dilution, ab109497, Abcam, UK), anti-SLC6A2 (NET, 1:250 dilution, MBS540046, MyBioSource, USA), and anti-HSP60 (1:1000 dilution, 12165S, Cell Signaling Technology, USA). The chemical luminescent signals were measured by Azure c400 Gel imaging system (Azure Biosystems, Inc. USA). Protein expression was quantified by drawing a region-of-interest (ROI) using ImageJ free software.

Quantitative RT-PCR analysis

Ground heart tissue powders were collected and soaked in Trizol (Invitrogen, USA) and RNeasy Fibrous tissue mini kit (Qiagen, USA) was used to isolate total RNA from heart tissues. Genomic DNA was removed by DNase I (Qiagen), and RNA was reverse transcribed using an iScript kit (Bio-Rad, USA). The resulting cDNA was analyzed by quantitative RT-PCR (qPCR) using SYBR green master mix (Life Technologies, USA) on QuantStudio6 Real-Time PCR system (Life Technologies). mRNA levels were calculated by delta-delta CT method using the target gene (Fap) and reference genes (Rpl32, Tbp, Gapdh, and Actb). The full primer list is reported in Table S5.

Bulk RNA-seq library construction and data analysis

The libraries were sequenced with paired-end 50 bps on the NovaSeq 6000 Sequencer (Illumina, USA). The raw sequencing reads in BCL format was processed through bcl2fastq 2.20 (Illumina) for FASTQ conversion and demultiplexing. After trimming the adaptors with cutadapt (version 1.18; https://cutadapt.readthedocs.io/en/v1.18/), RNA reads were aligned and mapped to the GRCm39 mouse reference genome by STAR (version 2.5.2; https://github.com/alexdobin/STAR),⁴⁴ and transcriptome reconstruction was performed by Cufflinks (Version 2.1.1) (http://cole-trapnell-lab.github.io/cufflinks/). The abundance of transcripts was measured with Cufflinks using fragments per kilobase of transcript per million mapped reads (FPKM) as an output.^45,46 Raw read counts per gene were extracted using HTSeq-count version 0.11.2.⁴⁷ Gene expression profiles were constructed for differential expression, cluster, and principle component analyses with the DESeq2 package (https://bioconductor.org/packages/release/bioc/html/DESeq2.html).⁴⁸ For differential expression analysis, pairwise comparisons were performed between two or more groups using parametric tests where read counts follow a negative binomial distribution with a gene-specific dispersion parameter. Corrected p-values were calculated based on the Benjamini-Hochberg method to adjust for multiple testing.

For the differentially expressed genes (DEGs) analysis, p < 0.01 was used as the signifier of statistical significance, and Log2FC (FC, fold change) ≥ 0.55 and Log2FC ≤ -0.85 were used to distinguish upregulated (Up) and downregulated (Down) DEGs, respectively. The heat map was generated using R studio to compare DEGs between groups, and the volcano plot for the overall distribution of DEGs was analyzed using GraphPad Prism 9.0 (GraphPad Software, USA).

DAVID analysis and establishment of PPI networks

The database for annotation, visualization, and Integrated Discovery (DAVID) was used to group DEGs based on biological function (https://david.ncifcrf.gov/). The 1326 Upregulated genes and 1684 Downregulated genes were submitted for the Gene Ontology (GO) according to the biological process (BP) analysis. The heat map for the GO:BP data was generated using R studio. A protein-protein interaction (PPI) network was developed to identify the association between a target and related DEGs by utilizing the STRING database (http://string-db.org/).⁴⁹ GO terms and PPI networks with a p-value cutoff < 0.05 were regarded as significant.

Histopathology

The tissue was processed in alcohol and xylene and embedded in paraffin. Four transverse sections of the heart per mouse, including right and left ventricles, right and left auricles, and interventricular septum were sectioned at 5-µm thickness and stained with hematoxylin and eosin. Histopathological evaluation of the heart was performed by a board-certified veterinary pathologist. Hearts were evaluated on the basis of cardiomyocytes showing necrosis, degeneration (cytoplasmic vacuolization), and atrophy, leukocytic cell infiltrates, and interstitial fibrosis.

Formalin fixed sections of the heart were stained with Masson’s Trichrome to evaluate the presence of collagen in cardiac tissues. To determine the percentage of collagen in the heart, digital whole slide images of Masson’s trichrome-stained hearts were manually annotated and then classified pixels were evaluated with a random forest algorithm using QuPath (an open-source software for digital pathology image analysis accessed through: https://qupath.github.io/). Regions of collagen for this analysis included collagen fibrils between cardiomyocytes and around preexisting vasculature within cardiac musculature. Regions excluded for this analysis included preexisting collagen from great vessels, leaflet insertion bands, and pericardial connective tissue.

Immunohistochemistry

Formalin-fixed, paraffin-embedded sections were stained using an automated staining platform (Leica Bond RX, Leica Biosystems). Following deparaffinization and heat-induced epitope retrieval in a citrate buffer at pH 6.0, the primary antibody against TSPO, also known as peripheral-type benzodiazepine receptor (PBR; ab109497, Abcam, Waltham, MA), was applied at a dilution of 1:10000. A rabbit anti-goat secondary antibody (Cat. No. BA-5000, Vector Laboratories, Burlingame, CA) and a polymer detection system (DS9800, Novocastra Bond Polymer Refine Detection, Leica Biosystems) was then applied to the tissues. The chromogen used was 3,3’-diaminobenzidine tetrachloride (DAB) and the sections were counterstained with hematoxylin and examined by light microscopy. Positive immunoreactivity for TSPO was confirmed with internal mouse tissue array controls used to validate this immunoassay. A subset of tissues incubated with antibody diluents and secondary antibody only were used as negative controls for this assay. Images were acquired with an Olympus VS200 slide scanner (Olympus, Tokyo, Japan) with a 20x objective. Quantitative image analysis was performed by using the QuPath Pixel classifier module. A random forest algorithm was used for identifying pixels as TSPO-positive, TSPO-negative and background in cardiac sections. Region of interest and thresholding values were validated by a board-certified veterinary pathologist.

For CD11b immunohistochemistry, a heat-mediated antigen retrieval with citrate buffer (pH 6.0) was applied on deparaffinized cardiac sections, which were then incubated with a primary anti-CD11b antibody at a dilution of 1:4000 (ab133357, Abcam, USA). A goat anti-rabbit secondary antibody (Cat. No. BA-1000, Vector Laboratories) and a polymer detection system (DS9800, Novocastra Bond Polymer Refine Detection, Leica Biosystems) were then applied to the tissues. The chromogen was DAB, and the sections were counterstained with hematoxylin and examined by light microscopy.

In Situ Hybridization

Formalin-fixed, paraffin-embedded cardiac sections were incubated with the target probe designed to detect region 486 - 1588 of murine fibroblast activation protein (Fap) mRNA, NCBI Reference Sequence NM_007986.3 (RNAscope® LS 2.5 probe for murine FAP, #423888; Advanced Cell Diagnostics, Newark, CA). The target probe was validated on sections of murine skin and heart from mice. Slides were stained on an automated stainer (Leica Bond RX, Leica Biosystems) with RNAscope 2.5 LS Assay Reagent Kit-Red (322150, Advanced Cell Diagnostics) and Bond Polymer Refine Red Detection (DS9390, Leica Biosystems). Control probes detecting a validated positive housekeeping gene (mouse peptidylprolyl isomerase B, Ppib to confirm adequate RNA preservation and detection; 313918, Advanced Cell Diagnostics) and negative control, Bacillus subtilis dihydrodipicolinate reductase gene (dapB to confirm absence of nonspecific labeling; 312038, Advanced Cell Diagnostics) were used. Positive RNA hybridization was identified as discrete, punctate chromogenic red dots under bright field microscopy. Images were acquired with an Olympus VS200 slide scanner with a 40x objective. Quantitative image analysis of Fap hybridization was performed with QuPath using an algorithm for singleplex chromogenic RNAscope image analysis. Fap positive hybridization signal was classified as follows: 1 red dot / cell, 2 red dot / cell, and 3+ red dots / cell in each transverse section of the heart. An H-score of Fap positive signal from each sample was calculated by the QuPath software. Samples with autolysis or regions in the tissue with pale-brown precipitate and/or folding artifacts were excluded from this analysis.

Statistical analysis

Statistical analyses were carried out using GraphPad Prism 9.0. All data were expressed as means ± standard deviation (SD) and are representative of at least three separate biological experiments. The unpaired two-tailed Student’s t-test or Mann-Whitney test was determined for comparisons of two groups. For correlation analysis, the Pearson correlation test was used. A p-value of less than 0.05 was considered statistically significant.

RESULTS

DOX treatment in mice induces cardiotoxic physiological and molecular changes

To establish a clinically-relevant model of DOX-induced cardiotoxicity, we administered intraperitoneal saline (control; CTRL) or DOX (cumulative dose 24 mg/kg) over 2 weeks (W)³⁹ to C57BL/6 mice followed by 10-12 W observation together with serial echo and microPET/CT imaging (Figure 1A). In agreement with multiple literature reports,^11,24 we observed significantly lower body weights in the DOX mice compared to the age-matched CTRL group (Figure 1B; Table S1). The heart weight (HW) to tibia length (TL) ratio was 40% lower in DOX groups compared to the CTRL group at from 7 to 12 W (p < 0.0001) (Figure 1C; Table S2). There were not significant differences in TL between the two groups (p = 0.5566) (Figure S1A), indicating that cardiac atrophy was occurring in the DOX animals, as previously reported.¹¹ Next, to evaluate the effect of DOX treatment at a cellular level, the expression of TOP2β, a primary mediator of DOX-induced toxicity,⁵⁰ was evaluated. We observed rapid and sustained decrease in TOP2β expression, which persisted up to 10 W (Figure 1D). Collectively, these experiments confirmed that a cumulative dose of 24 mg/kg DOX induced sustained cardiac atrophy and reduction of TOP2β protein expression in mice.

Fig. 1. Establishment of clinically-relevant cardiotoxic mouse model induced by systemic administration of DOX.

(A) Schematic of this study. After establishing the DOX-induced cardiotoxicity model, underlying pathophysiology was evaluated by serial PET imaging using [⁶⁸Ga]Ga-FAPI-04 (for targeting FAP), [¹⁸F]DPA-714 (for targeting TSPO), and [¹⁸F]MFBG (for targeting NET). Echocardiography (echo) imaging was performed at 4 weeks and 10 weeks. Heart tissue was extracted at the PET imaging time points for measurement of heart weight and evaluation of molecular and cellular changes in cardiac tissue. (B) Body weights were compared between CTRL and DOX groups over the course of the experiment. (C) Heart weight indexed to tibia length. (D) TOP2β expression from the heart lysates was evaluated by western blot. HSP60 was used as an internal control. Data are presented as the mean ± s.d. *** p < 0.001 and **** p < 0.0001.

[⁶⁸Ga]Ga-FAPI-04 PET detects cardiac abnormalities earlier than echo

We performed echo imaging and analysis in CTRL and DOX groups at acute (4 W) and chronic (10 W) phases.²⁴ At 4 W, left-ventricle end-diastolic diameter (LVDd) and left-ventricle end-systolic diameter (LVDs) were not significantly different between the CTRL (n = 6) and DOX (n = 6) animals (LVDd; p = 0.833, LVDs; p = 0.165) (Figure 2A-C). However, LVDs was significantly increased by 20% in the DOX group (n = 5) at 10 W (p < 0.01) (Figure 2A and C). In parallel, we observed no significant change in LVDd (p = 0.149) (Figure 2B). As a result, fractional shortening (FS) decreased from 40% in the CTRL animals and DOX animals at 4 W to less than 30% at 10 W (p < 0.01) (Figure 2D). At this later time point, blood cardiac troponin-I (CTNI) levels were higher in DOX mice, but the difference was not significant (p = 0.247) (Figure S2).

Fig. 2. [⁶⁸Ga]Ga-FAPI-04 PET detects cardiac remodeling before it is evident by conventional echo.

Pathology was monitored by echo and microPET/CT imaging. DOX-treated mice were randomly assigned to imaging groups for comparison with age-matched controls (CTRL). (A) Representative images of the left ventricle (LV) serial echo of CTRL (n =7) and DOX groups (4 W, n = 7 and 10 W, n = 5). Left ventricular end-diastolic diameter (LVDd) (B) and left ventricular end-systolic diameter (LVDs) (C) were determined from the echo scans. (D) Fractional shortening (FS) was calculated from LVDd and LVDs. Representative [⁶⁸Ga]Ga-FAPI-04 (E), [¹⁸F]DPA-714 (F), and [¹⁸F]MFBG (G) microPET/CT fusion transaxial images at 2 and 10 W. Quantitative [⁶⁸Ga]Ga-FAPI-04, [¹⁸F]DPA-714, and [¹⁸F]MFBG PET signals (mean value of % injected dose per centimeter cubed, %ID/cm³) was normalized by thigh muscle uptake. The number of mice imaged at each time point is displayed in Table S3. Data are presented as the mean ± s.d. * p < 0.05, and ** p < 0.01.

We performed serial microPET/CT imaging with [⁶⁸Ga]Ga-FAPI-04, [¹⁸F]DPA-714, and [¹⁸F]MFBG for 12 W to compare the time course of cardiac uptake differences with the time course of functional deficits in the DOX hearts (Figure S3; Table S3). To account for differences in heart function, we normalized cardiac uptake to skeletal muscle (thigh muscle). Skeletal muscle has minimal basal expression of our molecular targets and therefore acts as a surrogate for blood pool effects. Furthermore, it is also subject to DOX-induced toxicity¹¹ and therefore controls for off-target effects. [⁶⁸Ga]Ga-FAPI-04 uptake was significantly increased in the acute phase (2 W; 1.7-fold) (p < 0.01) when no evidence of cardiotoxicity is evident as measured by echo (Figure 2E). Moreover, increased cardiac [⁶⁸Ga]Ga-FAPI-04 uptake persisted in the DOX mice through the chronic phase (10 W; 1.5-fold) (p < 0.05) (Figure 2E, Figure S3). By contrast, there was no significant difference between cardiac uptake of [¹⁸F]DPA-714 and [¹⁸F]MFBG in the CTRL and DOX groups at either the early or late phases (Figure 2F and G, Figure S3).

Increased cardiac [⁶⁸Ga]Ga-FAPI-04 uptake is significantly correlated to FAP expression at the gene and protein levels

Next, we sought to validate the significant differences in cardiac [⁶⁸Ga]Ga-FAPI-04 uptake through determining their correlation with FAP protein and mRNA expression. Cardiac tissue was collected from perfused hearts at 4 W for Western blot, qPCR, and RNA-seq analyses and tissue staining (Figure 3A). As expected, FAP expression was 2.9-fold higher (p < 0.01) in DOX mice than CTRL animals, while TSPO and NET showed no significant differences in protein expression (Figure 3B and C). At the same time, we determined the Fap gene expression using three different Fap primers and different reference genes to ensure accurate qPCR analysis in spite of the heterogeneity of our heart tissue (Table S4).⁵¹ Fap gene expression increased 2.5-fold (normalized to Rpl32, p < 0.0001), 2.0-fold (normalized to Tbp, p < 0.0001), and 4.7-fold (normalized to Gapdh, p < 0.0001), respectively, compared with CTRL animals (Figure 3D). In agreement with our [⁶⁸Ga]Ga-FAPI-04 PET imaging, cardiac Fap gene expression in DOX mice increased 1.7-fold (p < 0.01), 2.8-fold (p < 0.0001), and 2.8-fold (p < 0.0001) relative to CTRL mice at 2, 7, and 10 W, respectively (Figure S4A). We observed a similar trend in FAP protein expression (Figure S4B), suggesting that mRNA and protein expression levels are proportional in this model. Additionally, FAP activity was significantly increased in DOX hearts compared to controls (p = 0.031) (Figure S5).

Fig. 3. Expression of FAP protein and mRNA is elevated at 4 W.

(A) Isolation of heart tissues for determination of mRNA, protein, and tissue levels. (B) Protein expression of FAP, TSPO, and NET (CTRL; n = 3, DOX; n = 4) was evaluated by western blot. HSP60 was used as an internal control. (C) Each band size was normalized by drawing ROIs using ImageJ free software. (D) RT-qPCR (CTRL; n = 3, DOX; n = 3) was performed for the validation of Fap mRNA expression using three different primers, listed in Table S5. Rpl32, Tbp, and Gapdh were used as reference genes in cardiac tissue. Data are presented as the mean ± s.d. ** p < 0.01 and **** p < 0.0001. The schematic illustration (A) was drawn using https://biorender.com/

FAP is a diagnostic imaging biomarker for detecting incipient cardiotoxicity by PET

Having established that the cardiac PET signals of our candidate probes correlated with protein and mRNA expression of the corresponding molecular target, we sought to validate uptake by tissue staining. We separated a cohort of DOX animals (n = 9) into high or moderate uptake groups (Figure 4A). Cardiac [⁶⁸Ga]Ga-FAPI-04 PET signal was significantly higher in these animals than in the age-matched CTRL mice (n = 4; p < 0.05). Interestingly, DOX-treated mice showed little evidence of pathological changes in cardiomyocytes. A mild degree of individual cardiomyocyte necrosis and degeneration and/or focal to multifocal aeas of myocardial fibrosis were occasionally observed in four DOX hearts. The rest of the DOX hearts (n = 11) did not show pathological changes.

Fig. 4. Correlation between FAPI PET, FAP in situ hybridization (ISH), and trichrome staining.

(A) Left: Representative whole-body FAPI PET sagittal and coronal maximum intensity projections (MIPs) acquired 45 min post injection of [⁶⁸Ga]Ga-FAPI-04. DOX mice showed moderate or high uptake compared to CTRL. Right: Image-based quantitation of cardiac PET signal normalized by thigh muscle uptake (H/M). (B) Representative H&E, Trichrome, and FAP ISH stains. Myocardial fibrosis was occasionally and regionally detected in DOX-treated hearts (arrows). Fap mRNA (punctate red dots) was detected in the cytoplasm and/or nuclei of cardiomyocytes and/or epicardial stromal cells (Insets). (C) Comparison of collagen positive regions from the trichrome staining. Quantitation was performed using QuPath software, with positive regions defined as deposition of collagen fibril between cardiomyocytes. (D) Comparison of the H-score from FAP ISH staining. Quantitation was performed using QuPath software. (E) A significant correlation was observed between [⁶⁸Ga]Ga-FAPI-04 cardiac PET signals and FAP expression in the corresponding tissues (p = 0.0003). (F) No correlation was observed between [⁶⁸Ga]Ga-FAPI-04 cardiac uptake and the percentage of collagen positive regions in the corresponding tissues (p = 0.3425). Data are presented as the mean ± s.d. * p < 0.05. Scale bar: 1 mm and 20 μm.

As DOX associated heart damage can lead to fibroblast activation and subsequent interstitial fibrosis, the percentage of collagen was evaluated by the use of Masson’s Trichrome staining of the hearts. There were no differences in H&E and Masson’s trichrome staining between these three groups (Figure 4B). Indeed, collagen-positive regions averaged approximately 7% for CTRL mouse hearts (n = 3) and approximately 6.5% for DOX hearts (n = 15) (Figure 4C). However, spatiotemporal Fap expression, as determined by in-situ hybridization (ISH), was significantly higher in the DOX animals (n = 11; p < 0.05) (Figure 4D). Fap nucleic acid was detected in the cytoplasm and associated with the nuclei of cardiomyocytes and stromal cells. Furthermore, the H-score was higher in the tissue slices belonging to mice in the high uptake group than mice in the moderate uptake group (Figure 4B; Table S6). There was a linear correlation between [⁶⁸Ga]Ga-FAPI-04 PET signal and H-score (p < 0.001) (Figure 4E) but no correlation between PET signal and collagen formation (p = 0.343) (Figure 4F). As predicted by our PET imaging, there was no difference in TSPO staining between DOX and CTRL mice (Figure S6). We observed no NET staining in either CTRL or DOX samples (data not shown). Taken together, these results indicate that FAPI PET is a potential diagnostic biomarker in the DOX model.

DOX promotes cardiac remodeling and disrupts mitochondrial energetics

To investigate the role that FAP may be playing in DOX-induced cardiotoxicity, we performed bulk RNA-seq analysis. We first constructed a volcano plot using Log2FC and a negative Log False discovery rate (FDR) with 14,101 DEGs. Although none of the Fap, Tspo, and Slc6a2 genes showed significant differences in the overall DEGs population, Fap gene expression did significantly increase when the expression level was normalized by fragments per kilobase of transcript per million (FPKM) mapped fragments (Figure S7A and B). We also generated a heat map of all DEGs. Clear differences between the DOX and CTRL groups were evident, though we observed a degree of heterogeneity within each group (Figure 5A). Compared with CTRL mice, DOX hearts showed 1326 markedly upregulated genes and 1684 markedly downregulated genes. These genes were used to identify the top 20 from the p-value affected biological processes (BP) in the gene ontology (GO). Within the upregulated genes, the most significantly affected GO:BPs included those related to intracellular signal transduction, protein phosphorylation, cell adhesion, angiogenesis, and extracellular matrix organization. Within the downregulated genes, on the other hand, the most affected GO:BPs were those related to mitochondrial translation, ATP synthesis, and respiratory chain complex (Figure 5B). Furthermore, numerous GO:BPs associated with cardiac remodeling were identified in the upregulated group, whereas GO:BPs associated with mitochondrial energetic dysfunction were identified in the downregulated group (Figure S8). Taken together, these data indicate that DOX treatment induces cardiac remodeling and impairs mitochondrial energetics in cardiomyoctyes.

Fig. 5. The apparent difference of biological processes in gene ontology according to DOX treatment.

(A) Heat map representation of the 14,101 genes from the same samples of qPCR. (B) The top 20 up-(red) or downregulated (blue) biological processes (BP) of gene ontology (GO) by P value were determined from the bulk RNA-seq data. (C) Heat map representation from the GO:Cell adhesion (upregulated; 26 genes) and GO:Mitochondrial ATP synthesis coupled proton synthesis (downregulated; 51 genes) with Fap (marked by #). (D and E) STRING database highlighting interactions between Fap and highly upregulated or downregulated genes.

The heat map was refined by selecting representative genes from upregulated or downregulated one BP of the top 3 GO:BP terms, including the Fap gene, and excluding genes with heterogeneous expression (Figure 5C). This heat map highlights the increased expression of Fap (indicated by the hashtag) in DOX hearts compared to CTRLs. Next, we analyzed the same gene family using the STRING database. FAP strongly interacted with fibronectin-1 (Fn1), a major component of cardiac ECM remodeling and fibrosis,^52,53 which in turn associated with other proteins involved in cell adhesion and migration (Figure 5D). On the other hand, there was no protein association between FAP and any component of the downregulated group (Figure 5E). From these findings, we conclude that increased FAP expression contributes to multiple processes involved in cardiac remodeling but does not predict fibrosis in cardiac tissue.

DISCUSSION

In this study, we sought to validate PET imaging targets that are more sensitive to the early symptoms of DOX-induced cardiotoxicity than conventional echo and are biomarkers of specific pathophysiologies. Nuclear medicine approaches have been underutilized for this purpose due to concerns about radiation exposure and lack of widespread availability of imaging devices in cancer treatment facilities. Nevertheless, this imaging technique allows anatomical information to be coupled to biochemical information afforded by a suitably chosen probe targeting a specific disease-relevant molecule or pathway. Although DOX-induced cardiotoxicity ultimately decreases LVEF, compromises cardiac performance, and can lead to heart failure, the mechanisms by which it does so may vary between patients. Non-invasive detection of these mechanism by molecular imaging is a safe and important way of assessing individual-to-individual variations. This is significant because the appropriate cardioprotective treatment could be administered on an individual basis with suitable knowledge of the specific underlying pathology. This is the premise behind personalized medicine.

The primary nuclear medicine approaches to imaging cardiotoxicity focus on imaging perfusion.⁵⁴ Recently, new probes have been developed to image specific pathophysiologies, including mitochondrial damage,^55,56 sympathetic inervation,^34,57–59 inflammation,^31,60–62 cardiac remodeling,^63–67 and cardiac metabolic dysfunction.^68–70 These probes have largely been studied in preclinical models, where they have provided important insight into the development and progression of cardiotoxicity. To build on this work, we identified cardiac remodeling, inflammation, and inervation as plausible contributors to DOX-induced cardiotoxicity that could arise at the earliest stages of disease. We selected three radioligands, [⁶⁸Ga]Ga-FAPI-04,^71,72 [¹⁸F]DPA-714,^35,73 and [¹⁸F]MFBG,⁷⁴ that are already under clinical investigation for other indications to image these processes. Preliminary reports indicate that the radiation dosimetry of these probes is acceptable, thereby supporting their use in cancer survivors. We anticipate that this will facilitate the future clinical translation of our PET approach.

To further accelerate clinical translation, we applied a well-established model of cardiotoxicity encompassing subacute and chronic phenotypes.^24,39 This heterogenous model reflects the reality that anthracycline-induced cardiotoxicity is a complex process that involves multiple cell types in heart tissue⁷⁵ and often differs between even those patients receiving the same dose of anthracycline. In cardiomyocytes with high TOP2β expression,⁷⁶ DOX binds to DNA form a complex with TOP2β that triggers cell death pathways.^5,50 Consequently, cancer patients with high levels of TOP2β in cardiomyocytes are likely to be more susceptible to DOX-induced cardiotoxicity.^50,77 Our Western blot data confirm that TOP2β expression is rapidly downregulated in response to DOX treatment, with gradual recovery over the 12 W observation period. Furthermore, we observed body weight decrease and diminishing HW/TL ratios after DOX treatment, consistent with established models of DOX-induced cardiotoxicity^11,78. Although not statistically significant at this dose, heart weight was a lower percentage of body weight in the DOX mice than the controls (Figure S1B), suggesting that cardiac atrophy is more pronounced than cachexia or other systemic effects in these animals.¹¹ Finally, functional declines in cardiac performance emerged by 10 W, even when corresponding interstitial fibrosis was not widely observed. These observations indicate a spectrum of symptoms consistent with clinical presentation of cardiotoxicity.

Our transcriptomic data indicate that cardiac remodeling is initiated in response to DOX treatment. Pathological cardiac remodeling arises in chronic heart failure through the activation of multiple pathways,^79,80 and we observed intracellular signaling,^81–83 cell adhesion,⁸⁴ angiogenesis,⁸⁵ extracellular matrix remodeling,^86,87 and cell migration,⁸⁸ to be highly enriched in our DOX tissues. Another major aspect of remodeling is cardiac fibrosis. This process is typically initiated by activated cardiac fibroblasts and eventually leads to the functional change of heart tissue and diastolic dysfunction.⁸⁹ FAP is a marker of activated cardiac fibroblasts⁸⁹ and has recently been targeted by radiopharmaceuticals for PET imaging in cardiovascular disease and heart failure.^65,90–92 In addition, a recent case study speculated that incidental cardiac FAPI PET signal detected in a cancer patient may have been due to cardiotoxicity arising from the chemotherapy regimen,⁷ although this hypothesis was not explored further. Furthermore, the BioGPS⁹³ and GTEx databases indicate basal Fap gene expression to be moderate in cardiac tissue of mice and humans (Figure S9), which may enable relatively small changes in expression to be detected by molecular imaging. These observations provided the rationale for our hypothesis that FAPI PET would detect incipient cardiotoxicity.

We demonstrate that the cardiac signal intensities of [⁶⁸Ga]Ga-FAPI-04 PET increased almost immediately after DOX treatment, substantially earlier than any functional alteration could be imaged by echo. This increased signal, which was sustained throughout the 12 W observation period, significantly correlated with expression of FAP protein and Fap mRNA. We were unable to successfully perform immunohistochemistry for murine FAP using commercially available antibodies, and therefore used ISH to assess the distribution of FAP in cardiac tissue. Our experiments confirm the specificity of FAPI PET for DOX-induced alterations in cardiac FAP, thereby highlighting the initiation of pathological remodeling pathways in the injured heart. We also provide a mechanistic link between elevated PET signal and disease, an outcome that has not been accomplished in human patients due to inconclusive histopathology studies.^65,66 Significantly, our data indicate that FAP is a diagnostic imaging biomarker in cardiotoxicity that might be superior to echo in detecting early mediators of cardiac damage.

Our study did not conclusively establish whether FAP is also a prognostic imaging biomarker in cardiotoxicity. We did not find a correlation between PET signal and fibrosis in our tissue samples. However, fibrosis was minimal in our model. We conclude that FAP in this model participates more broadly in cardiac remodeling. Consistent with literature reports in mice and humans,^11,94 DOX treatment induced cardiac atrophy in our mice. Cardiac atrophy requires extensive remodeling of the ECM due to loss of cardiomyocyte mass,⁹⁵ consistent with the upregulation of ECM remodeling pathways evident in our RNA-seq analysis (Figure 5B; Figure S7). We did not see evidence of substantial cardiomyocyte death, which may explain why we observed minimal fibrosis in this model. Indeed, reports of fibrosis in atrophied hearts are conflicting,⁹⁴ which likely reflects both the techniques used to quantify collagen and the prevalence of remodeling pathways that do not result in collagen deposition. Nevertheless, given that DOX induces cardiotoxicity through a variety of molecular mechanisms,⁹⁶ it is possible that FAP PET might correlate with other indices of disease severity. For example, we observed a negative correlation between cardiac FAPI PET signal and HW/TL at the end of the study (Figure S10). We will need larger group sizes, longer follow up periods, alternative indices of disease severity, and perhaps more acute pathology to determine if elevated cardiac FAPI PET signal corresponds to more severe long-term outcomes.

Our alternative molecular targets, TSPO and NET, proved to be neither diagnostic nor prognostic biomarkers in DOX-induced cardiotoxicity. Our rationale for targeting TSPO was the prominent role that oxidative stress and inflammation play in DOX-induced cardiotoxicity.³⁷ TSPO is expressed not only in cardiomyocytes⁹⁷ but also in activated immune cells, especially macrophages.^98–100 Resident and circulating macrophages are implicated in the response to DOX-induced cardiotoxicity.¹⁰¹ Preclinical studies have demonstrated increased cardiac uptake of [¹⁸F]DPA-714 in mice with inflammatory heart conditions,^60,102 but we observed neither a significant increase in [¹⁸F]DPA-714 signal nor an increase in TSPO staining by immunohistochemistry. Moreover, we found no evidence of macrophage infiltration by histology or CD11b immunohistochemistry (Figure S11). TSPO is also present in the mitochondrial outer membrane of cardiomyocytes where it modulates oxidative stress and regulates mitochondrial physiology and metabolism.¹⁰³ We observed downregulation of a number of genes related to mitochondrial metabolism in the DOX mice, but this did not translate to increased [¹⁸F]DPA-714 uptake. Given the nearly ubiquitous expression of TSPO in tissue,¹⁰⁴ it is possible that substantial off-target uptake reduces the sensitivity of the radioligand for changes in cardiac expression induced by DOX. Moreover, its high basal expression in human and murine heart (Figure S9), may render TSPO imaging insensitive to small changes in expression levels. Additionally, the systemic inflammation induced by DOX treatment in this mouse model may further obscure small changes in cardiac PET signal and may represent a limitation of our model.

Our decision to target NET with [¹⁸F]MFBG was based on prior evidence that cardiac uptake of radiolabeled meta-iodobenzylguanidine (MIBG) decreases in a dose-dependent manner in rodents treated with anthracyclines⁵⁷ and in cancer patients that had received anthracycline chemotherapy relative to those receiving alternative treatment.^38,105 We saw considerable cardiac uptake normalized to skeletal muscle of [¹⁸F]MFBG in both DOX animals and CTRL (Figure S3), but no decline over the 12 W observation period. By contrast, in the early stages of cardiotoxicity, [¹⁸F]MFBG uptake was actually higher in DOX mice, though this was not statistically significant (1 W; p = 0.400, 2 W; p = 0.527, and 4 W; p = 0.161). Although prior studies did show declines in [^123/125I]MIBG uptake concurrent with LVEF decline, the uptake deficit was sustained. It may be that our follow up period was too short to detect differences between our groups, but the convergence of the curves in Figure S3 suggests that differences are unlikely to emerge. A retrospective analysis could not discriminate asymptomatic pediatric cancer survivors from healthy controls using [¹²³I]MIBG image quantification, and myocardial sympathetic activity was neither related to anthracycline dose nor LVEF.¹⁰⁶ This may indicate that sympathetic denervation is not sufficiently pronounced in chronic cardiotoxicity to represent a reliable imaging biomarker.

To date, echo has been used in cardio-oncology as the main imaging modality for screening patients with suspected cardiotoxicity.¹⁰⁷ Given the implementation of new echo techniques that improve its sensitivity for subclinical disease and the continued definition of cardiotoxicity in terms of LVEF decreases, echo will continue to play a major role in diagnosis and monitoring progression. However, our results support a role for PET imaging in the management of cancer patients receiving anthracycline chemotherapy. In our model, we detected pathological cardiac remodeling in DOX hearts as much as 8 weeks before functional decline was evident by echo. Early diagnosis of cardiotoxicity could greatly improve its treatment, as evidenced by the more complete recovery of LVEF in patients with cardiotoxicity administered ACE inhibitors or beta blockers shortly after anthracycline chemotherapy than patients treated a few months later.¹⁰⁸ Molecular imaging techniques such as FAPI PET may lead to even more impressive treatment outcomes by identifying the activation of specific pathological pathways whose inhibition could mitigate or even prevent cardiotoxicity. For example FAP inhibition improves cardiac repair after myocardial infarction.^109,110 Future work is needed determine whether it is similarly beneficial in cardiotoxicity, but this example does highlight the potential benefit of PET imaging biomarkers in treating cardiotoxicity.

We acknowledge several limitations of our study. Firstly, although we showed the correlation between cardiac FAPI PET uptake and FAP expression, elevated background FAPI PET signal in DOX mice was also seen due to uptake in the gastrointestinal region, muscle, and in some cases, lung. This likely reflects off-target uptake due to sustained and global inflammation caused by systemic administration of DOX. This phenomenon was previously observed in FAPI PET imaging of a pre-clinical model of idiopathic pulmonary fibrosis induced by bleomycin.¹¹¹ Secondly, our methods of quantifying FAP protein expression could not distinguish between membrane-bound FAP and cytoplasmic FAP. As our radioligand does not cross the cell membrane, the signals derived from this probe reflect the binding of membrane-bound FAP. To our knowledge, FAP is primarily an outer membrane protein, though increased cytoplasmic expression was recently reported in lung adenocarcinoma cells.¹¹² We therefore cannot rule out the possibility that cytoplasmic FAP protein expression confounds our analysis even though our results identify increased FAP protein expression and activity and gene expression in the DOX mice. Moreover, our studies have not determined the function of FAP in DOX-induced cardiotoxicity. Contrary to our expectations, increased FAP expression did not result in increased fibrosis. We speculate that FAP is broadly involved in ECM remodeling, but without identifying its specific role in this pathology, it will be challenging to determine whether FAPI PET could also be a prognostic biomarker in cardiotoxicity. Larger sample sizes could possibly determine whether early increases in FAPI PET correspond to larger declines in functional parameters such as LVEF. Finally, our studies were limited to male mice because female mice are less susceptible to cardiotoxicity.¹¹ Therefore, further research is required to determine whether FAPI PET will be equally valuable in female patients.

CONCLUSIONS

Although anthracycline chemotherapy has dramatically improved treatment outcome in cancer patients, especially in children with cancer, it causes cardiotoxicity with an increased risk of heart failure in a significant number of patients. Existing imaging modalities detect cardiac functional deficits but do not identify the underlying, potentially treatable, pathologies responsible for these deficits. We demonstrate a significant and sustained increase of FAP expression in response to systemic administration of doxorubicin and show that this change can be imaged by PET using [⁶⁸Ga]Ga-FAPI-04. Functional changes were not evident by routine echo until 10 weeks, as much as 8 weeks after cardiac FAPI PET signal increases were detected. These findings suggest that FAPI PET is a diagnostic imaging biomarker for incipient cardiotoxicity and a potential complement to echo for the management of cancer patients receiving anthracycline chemotherapy. Early detection of FAP-mediated cardiac remodeling may improve the efficacy of therapeutic interventions to delay or even prevent heart failure.

SOURCES OF FUNDING

This work was partially funded by award R21CA246409-01 (J.M.K.) from the National Cancer Institute (NCI) of the National Institutes of Health and by NCI Cancer Center Support Grant P30CA008748 issued to Memorial-Sloan Kettering Cancer Center (Laboratory of Comparative Pathology, S.E.C.). The funding agency did not influence the design, execution, or interpretation of the experiments.

DISCLOSURES

J.M.K. and J.W.B. hold intellectual property rights for compounds targeting FAP.

AUTHOR CONTRIBUTIONS

The studies were conceived by J.M.K., J.W.B., and A.diL. Experiments were designed by C-H.L. and J.M.K. and performed by C-H.L, O.M., L.R., and T.M.J. Data analysis was performed by C-H.L., O.M., S. Cho., T.M.J., A. diL., and J.M.K. Pathology and quantitation of ISH and IHC was performed by S.E.C. Funding was acquired by J.M.K. The manuscript was written by C-H.L. and J.M.K. and reviewed by O.M., L.R., S.E.C., S. Cho., T.M.J., J.W.B., and A. diL.

Nonstandard Abbreviations and Acronyms

ACE

Acetylcholinesterase

BP

Biological process

BSA

Bovine serum albumin

CT

Computed tomography

CTRL

Control

DEG

Differentially expressed genes

DOX

Doxorubicin

FAP

Fibroblast activation protein alpha

FAPI

Fibroblast activation protein alpha inhibitor

FC

Fold change

FPKM

Fragments per kilobase of transcript per million mapped reads FS Fractional shortening

GO

Gene oncology

HW

Heart weight

ID

Injected dose

ISH

In situ hybridization

LVDd

Left ventricle end-diastolic diameter

LVDs

Left ventricle end-systolic diameter

LVEF

Left ventricular ejection fraction

MFBG

meta-Fluorobenzylguanidine

MIBG

meta-Iodobenzylguanidine

MIP

Maximum intensity projection

NET

Norepinephrine transporter

PBR

Peripheral-type benzodiazepine receptor

PBS

Phosphate buffered saline

PET

Positron emission tomography

RNA-seq

NA sequencing

ROI

Region of interest

SPECT

Single photon computed tomography TL Tibia length

Top2β

Topoisomerase-2β

TSPO

Translocator protein 18-kDa

VOI

Volume of interest

WB

Western blot

ACKNOWLEDGMENTS

The authors wish to acknowledge Anil Ekkati, Ph.D. and David J. Warren, Ph.D. of the Milstein Core Chemistry facility for synthesizing DPA-714 and its tosylate precursor, and Pradeep K. Singh, Ph.D., and Stephen G. DiMagno, Ph.D. of the University of Illinois-Chicago for support with ALP-mFBG synthesis. The authors also wish to acknowledge Alejandro Amor-Coarasa, Ph.D., of Ratio Therapeutics, Serge Lyashchenko, Pharm.D., and Eva Burnazi of Memorial-Sloan Kettering Cancer Center for helpful discussions concerning the radiosynthesis of [¹⁸F]MFBG.