CXCR4 Is a Potential Target for Diagnostic PET/CT Imaging in Barrett's Dysplasia and Esophageal Adenocarcinoma

This article is featured in Highlights of This Issue, p. 987

The chemokine receptor 4 (CXCR4) has been shown to be consistently overexpressed in a variety of solid tumors (1–3). Its activation by the only endogenous ligand, CXCL12 (formerly termed SDF-1), contributes to cancer progression through both autocrine and paracrine stimulation of cancer growth, as well as inhibition of apoptosis in neoplastic cells (4). Alternatively, CXCR4 signaling may trigger tumor angiogenesis by regulating proangiogenesis factors (5) and attracting endothelial cells to the tumor microenvironment (6). Furthermore, CXCR4 overexpression promotes tumor invasiveness and metastasis by facilitating retention and homing of tumor cells in cellular niches such as the bone marrow. CXCR4 is normally expressed by myeloid cells, B cells, and T cells (7, 8), but is also thought to be expressed by some epithelial cells. In the stomach, CXCR4 also identifies type 2 innate lymphoid cells (ILC2; ref. 9). Expression of CXCR4 by hematopoietic progenitor cells (HPCs) and the production of CXCL12 by stromal cells and osteoblasts are important for homing and retention of these cells. Together with other cytokines (e.g., SCF and IL7), CXCR4 function is required for normal maturation of myeloid and lymphoid cells, and for the survival and proliferation of B-cell precursors and myeloid progenitor cells (10). While CXCR4 has also been shown to be expressed by columnar epithelial cells in the gastrointestinal tract, where it appears to mark progenitors, the role of CXCR4 signaling in this setting remains to be elucidated (11).

The incidence of esophageal adenocarcinoma, a highly invasive cancer, is rapidly rising in the Western world, and now accounts for 2% of all cancer-related deaths (12). CXCL12 (SDF-1) and CXCR4 expression in esophageal adenocarcinoma are associated with a poor prognosis and metastases to lymph nodes or bone marrow (13, 14). In vitro, CXCR4-positive (CXCR4⁺) esophageal cancer cells show stronger migration ability (15). Knockdown of CXCR4 expression by siRNA or the use of inhibitors reduces the proliferation and invasion ability of esophageal cancer cells (4, 16). In vivo, pharmacologic antagonism of CXCR4 has been shown to decrease tumor growth and metastasis in xenograft models (4, 16, 17). However, these studies have limitations in their ability to accurately reflect human disease, and do not address the role of CXCR4 in the premalignant stage of Barrett's esophagus. Barrett's esophagus is the predominant precursor lesion for esophageal adenocarcinoma, and is strongly associated with gastroesophageal reflux disease (GERD); although it remains unclear why some patients with GERD develop Barrett's esophagus and others do not (18). Furthermore, the overall rate of cancer progression from Barrett's esophagus is extremely low, and thus specific biomarkers are needed to identify patients with an elevated risk for developing esophageal adenocarcinoma. Detecting cancer at an early and potentially curable stage in early Barrett's esophagus progression might considerably improve survival of patients with esophageal adenocarcinoma. In this study, we investigated the role of CXCR4 expression in esophageal tumor progression, and explored its relevance for early tumor detection using molecular imaging.

We have previously developed the pL2-IL1β transgenic mouse model of Barrett's esophagus that exhibits chronic inflammation in the esophagus and recapitulates the histologic progression to esophageal adenocarcinoma (19). To date, this mouse model has provided several fundamental insights into the pathogenesis of Barrett's esophagus, including the emerging paradigm that Barrett's esophagus and esophageal adenocarcinoma arise from distinct progenitor cells in the gastric cardia (19). Considerable evidence supports chronic inflammation as a major factor in the development of Barrett's esophagus and esophageal adenocarcinoma, as it is able to trigger the expansion of these gastric cardia progenitors. In some patients with GERD, chronic injury to the GE junction resulting from exposure to gastric contents leads to the upregulation of inflammatory cytokines (IL1β, IL6, IL8), which are thought to contribute to Barrett's esophagus and esophageal adenocarcinoma. IL1β is highly expressed in Barrett's esophagus (20), and the pL2-IL1β mouse carries a modified human IL1β transgene controlled by the Epstein-Barr virus (L2) promoter, resulting in increased IL1β gene expression targeting the esophagus and squamous forestomach. The pL2-IL1β mice exhibit chronic esophagitis and progress to Barrett's esophagus by 6 months, eventually developing esophageal adenocarcinoma at an older age.

Here, we show that CXCR4 expression increases in both immune and epithelial cells during progression of Barrett's esophagus to low-grade (LGD) and high-grade dysplasia (HGD) and ultimately to esophageal adenocarcinoma in the pL2-IL1β mouse model. We find that CXCR4⁺ immune cell recruitment specifically correlates with increased dysplasia progression, suggesting that immune cells are key contributors to esophageal carcinogenesis. We also find that CXCR4 marks epithelial cells near the base of gastric cardia glands, which expand during the growth of these cells into the esophagus. Enhanced CXCR4-related signal intensity in ex vivo fluorescence imaging and autoradiography were found to correlate with increased CXCR4 expression during disease progression in this model. A novel technique for CXCR4-directed [⁶⁸Ga]pentixafor-PET/CT imaging in patients with esophageal adenocarcinoma demonstrates focal tracer accumulation in esophageal cancer and in metastatic sites. Our results support the notion of exploiting CXCR4⁺ immune cells together with CXCR4⁺ epithelial cells for molecular imaging of the tumor microenvironment, thus identifying CXCR4 as a promising biomarker for monitoring and staging of esophageal tumor progression.

A genetic mouse model (pL2-IL1β) that overexpresses IL1β in the mouse esophagus and stomach was prepared as described previously (19). Mice were allowed a standard chow diet from birth until weaning, and water ad libitum. Following weaning and genotyping, between 6 and 8 weeks of age, mice were assigned high-fat diet or remained on the chow diet: Ssniff, V1124-000, Metabolizable energy (ME):61 kJ%% energy from carbohydrate, 27 kJ% from protein and 12 kJ% from fat). High-fat diet: (Ssniff, S5745-E712, ME: 34 kJ% energy from carbohydrate, 18 kJ% from protein and 48 kJ%% from fat). All animal experiments were approved by the District Government of Upper Bavaria and performed in accordance with the German Animal Welfare and Ethical Guidelines of the Klinikum rechts der Isar, TUM (Munich, Germany).

For mouse samples, after sacrifice, stomach and esophagus tissues were fixed in formalin and paraffin embedded. Tissues were cut and put on superfrost slides for drying overnight or at the 60°C oven for >1 hour. Standard IHC procedures were performed using following antibodies: rat anti-mouse CD184 (CXCR4) antibody (eBioscience, 1:250 4°C overnight) and rabbit anti-α-SMA (Abcam, 1:400, 2 hours room temperature) for mouse tissue. Quantification of CXCR4 was assessed on three serial sections (2–3 μm) taken every >100 μm for each mouse. Five high-power fields [field of vision under maximum magnification (×400)] were taken by microscope (Zeiss) in each section. The images were taken in a manner to achieve high power fields with the most positively stained cells. Then CXCR4⁺ cells (brown) and hematoxylin-counterstained nuclei of the cells (blue) were counted. CD31 immunofluorescence staining was done as described previously (9). For human biopsy samples, Barrett's esophagus, HGD, and adenocarcinoma tissues from esophagectomy or endoscopic resection specimens were identified from archived formalin-fixed and paraffin-embedded (FFPE) cases from the Department of Pathology and Cell Biology, Columbia University (New York, NY). Five-μm-thick tissue sections were deparaffinazed and rehydrated, followed by antigen retrieval, performed in 100-mL Target Retrieval Solution pH9 (Dako, #S2367) using a steaming approach for 25 minutes followed by cooling at room temperature for 20 minutes; then, slides were rinsed three times with PBS for 5 minutes. Endogenous peroxidase was blocked in 30% hydrogen peroxide for 10 minutes. The slides were rinsed with distilled water once and with PBS three times for 5 minutes each time. The slides were then blocked in 10% normal goat serum for 20 minutes at room temperature. After removal of blocking serum, the sections were incubated with anti-CXRC4 antibody (Abcam, #ab124824) at 1:1,000 dilution for 90 minutes at room temperature. Staining of primary antibody was detected using anti-rabbit biotinylated secondary antibody (Vector Laboratories, # BA-1000) at 1:300 dilution for 30 minutes at room temperature followed by avidin-biotinylated peroxidase solution for 30 minutes at room temperature. DAB solution (Dako, #K3468) was used as the chromogen; the sections were then counterstained with hematoxylin (Merck Millipore, #105174).

Tissue microarrays from patients with esophageal adenocarcinoma were constructed from formalin-fixed paraffin-embedded (FFPE) tissue blocks obtained from 47 patients with esophageal adenocarcinoma diagnosed at the Charité University Hospital (Berlin, Germany) and 12 patients diagnosed at the University of Heidelberg (Heidelberg, Germany). Patient material was accessed after written consent and according to the regulations and ethical vote of the Technical University of Munich (503/16s). IHC for CXCR4 was performed with a Leica Bond Rxm system (Leica) with a primary CXCR4 antibody (Abcam, #ab124824) and a polymer refine detection system.

Single-cell suspensions of murine esophageal and cardia tissue along with forestomach regions were generated by chopping tissue with scissors in EDTA solution. Then, the tissue and EDTA solution was transferred into digestion medium followed by incubation at 150 rpm at 37°C for 30 minutes. Digestion medium for esophagueal tissue consisted of 5 mL Krebs Ringer buffer +4% (w/v) BSA (0.2 g) + 2 mg/mL collagenase P (Roche). Digestion buffer for cardia and forestomach consisted of 5 mL DMEM + 2 mg/mL collagenase P + 2 mg/mL Pronase (Roche). The following antibodies were used for FACS analysis: e450-CD45, APC e780-cd11b, APC- F480, Alexo700-Ly6G, FITC- cd3, and PE-anti-CXCR4, APC e780-Epcam, APC e780-NK1.1, PECy5.5-B220, and PECy7-CD90.2 (all antibodies were purchased from eBioscience). 7-AAD (eBioscience) was used to quantify live cells. FACS data were acquired on a Gallios flow cytometer (Beckman Coulter) and analyzed using FlowJo software (TreeStar).

CXCR4 antibody (eBioscience) was conjugated with Cy5.5 (Alexa Fluor 680, AF680)-NHS Ester (Life Technologies) according to the manufacturer's protocol. Mice were injected intravenously with conjugated CXCR4-Cy5.5 antibody and sacrificed 24 hours postinjection. After sacrificing, the esophagus and stomach were excised and imaged. Fluorescence images were acquired by illuminating the specimens using 670- and 750-nm diode lasers and guiding the emitted fluorescence through appropriate emission filters before capturing it using a back illuminated EM-CCD camera (iXon DU888, Andor), as described previously (21). Signal specificity was determined by ImageJ for calculating the ratio of the mean signal intensity in the exposed esophagus tissue and the surrounding background. To demonstrate the specificity of the fluorescent probe, mice were coinjected with CXCR4-Cy5.5 and the CXCR4 inhibitor AMD3100 (2 mg/kg). Injection of AMD3100 resulted in a near absence of significant signals in early lesion regions at the SCJ and in esophagus, supporting the notion that the fluorescent probe had binding specificity for CXCR4 (data not shown).

Mice (12 months treated with chow diet or high-fat diet, n = 2 per group) were injected intravenously with approximately 1.85 MBq [¹²⁵I]TUM4007 in 100-μL PBS (pH 7.4) under isoflurane anesthesia. Like pentixafor, [125I]TUM-4007 is not an antibody, but is a peptidic probe with a molecular weight <2,000 Da. After 1 hour, mice were sacrificed and perfused with 4% PFA. The tissues of interest were resected and stored at (−80°C). Cryosections (5 μm) of stomach/esophagus were then mounted on cover slides and exposed on a Fujifilm BAS-IP MS 2025 (20 × 25 cm) imaging plate (GE Healthcare Lifesciences). The plate was scanned with a CR35 Bio Phosphor Imager (Raytest) in sensitive 25-μm resolution mode. Scanning and data export was performed using AIDA software (Raytest).

[⁶⁸Ga]pentixafor-PET/CT was performed on five patients (one woman, four men; mean age, 72 ± 12.1 years). Overall, nine PET/CT examinations were carried out. The validation of images and SUV values were performed by an experienced nuclear medicine consultant. [⁶⁸Ga]pentixafor was used in six and [¹⁸F]FDG in three scans. In three patients, both [⁶⁸Ga]pentixafor- and [¹⁸F]FDG-PET/CT imaging was performed at the same time. One patient received [⁶⁸Ga]pentixafor-PET/CT imaging before and after chemotherapy. Another patient only received [⁶⁸Ga]pentixafor-PET/CT. Injected activity for [⁶⁸Ga]pentixafor was 163 ± 52 MBq and 361 ± 61 MBq for [18F]-FDG. The waiting time between injection and imaging was 56 ± 9 minutes for [⁶⁸Ga]pentixafor and 65 ± 6 minutes for [¹⁸F]FDG. The CT scan protocol included a low-dose CT from the base of the skull to the mid-thigh for attenuation correction followed by the PET scan and in some cases a diagnostic CT in the portal venous phase. PET scans were taken with five to seven bed positions for 2 to 3 minutes each. PET data were reconstructed with 128 × 128 pixel images. Images were reconstructed by an attenuation-weighted ordered subsets expectation maximization algorithm (four iterations, eight subsets) followed by a postreconstruction smoothing Gaussian filter (5-mm full-width at half-maximum). For semiquantitative SUV mean evaluation, a 3D volume of interest (VOI) with a growing seeded method of a 50% isocontour of SUV_max was used.

We utilized our transgenic mouse model (pL2-IL1β) of Barrett's esophagus and dysplasia to analyze the pattern of CXCR4 expression during esophageal carcinogenesis (19). pL2-IL1β mice develop low-grade dysplastic lesions in the setting of Barrett-like metaplasia at the squamocolumnar junction (SCJ) at approximately 12 months of age. When pL2-IL1β mice were fed a high-fat diet (HFD), they develop accelerated disease, with mid-grade dysplastic lesions along the SCJ by 9 months and larger high-grade dysplastic lesions in the esophagus and SCJ at 12 months (Fig. 1A, i and ii), enabling us to study changes associated with stepwise progression of esophageal carcinogenesis. CXCR4⁺ cells were detected by IHC only in the inflamed esophagus of pL2-IL1β mice, but not the esophagus of wild-type mice (Fig. 1A, iii). Notably, pL2-IL1β mice showed increasing CXCR4 expression levels in the esophagus and at the SCJ, which correlated well with increasing numbers of macroscopic and microscopic dysplastic lesions, along with increasing age (Fig. 1A and B). Furthermore, significantly higher CXCR4 expression was found in tissue samples of human esophageal adenocarcinoma compared with human Barrett's esophagus tissue with only LGD (Fig. 1C and D). Overall, these data suggest that CXCR4 expression correlates with esophageal tumor progression from nondysplastic Barrett's esophagus to LGD to HGD and esophageal adenocarcinoma.

Interestingly, our IHC analysis showed that the majority of CXCR4 expression was in the stroma, suggesting that esophageal epithelial (Fig. 1C, solid arrows) or cancer cells may not be solely responsible for the observed increases in CXCR4 expression. Therefore, we next investigated the nature of the cells contributing to the accelerated expression of CXCR4 in dysplastic esophageal tissue. IHC showed that a-SMA⁺ fibroblast cells in the esophagus did not express CXCR4 (Fig. 2A), indicating that the higher numbers of a-SMA fibroblasts were also not the main contributors to elevated CXCR4 expression during esophageal carcinogenesis, as previously suggested in other cancer models (6, 22). In addition, CD31⁺ cells did not show any overlap with the GFP reporter gene in a CXCR4-GFP reporter mouse crossed to the pL2-IL1β mouse model at different stages of carcinogenesis (Fig. 2B), indicating that endothelial cells are also not contributing substantially to the CXCR4⁺ cell pool.

In contrast, there were multiple CXCR4⁺ immune cells detectable in esophageal adenocarcinoma patient biopsies (Fig. 1C, opened arrows). Mice with both LGD and HGD, respectively, showed a similar abundance of CD45⁺ immune cells in the inflamed esophagus, indicating a significant inflammatory infiltration (Fig. 2C). Similar findings were also noted in patients with Barrett's esophagus and esophageal adenocarcinoma (Fig. 2C). Interestingly, in the esophagus of 9- to 12-month-old pL2-IL1β mice, we observed increasing numbers of CXCR4⁺ leukocytes (CXCR4⁺CD45⁺), particularly increased myeloid cell (CD45⁺CD11b⁺), neutrophil (CD45⁺CD11b⁺F480⁻Ly6G⁺) and CD3⁺ T-cell populations (Fig. 2C–G). There was no difference in B cells, ILCs, and NK-cell populations (Supplementary Fig. S1B–S1D). Moreover, we also observed significantly increased numbers of CXCR4⁺ myeloid cells in LGD lesions from 9 months to 12 months (pL2-IL1β mice on chow diet; Fig. 2I). Taken together, these data suggest that immune cells might be the main contributors to the elevated stromal CXCR4 expression, and are detectable at early dysplastic stages, comprising a distinct pretumor microenvironment. In addition, increased CXCR4⁺ myeloid cells were found in the spleen (Supplementary Fig. S1E), implying that the CXCR4⁺ myeloid cells in esophagus may be recruited in part from extramedullary sites. Together, these data suggest that these elevated CXCR4⁺ immune cell populations might be the main contributors to a procarcinogenic inflammatory niche that is distinct from routine esophageal inflammation, and instead leads to dysplasia and cancer.

Barrett's esophagus is defined as replacement of the stratified squamous epithelium in the distal esophagus with a metaplastic, intestinal-like columnar epithelium (23). In our murine model of Barrett's esophagus and esophageal adenocarcinoma, the metaplastic lesions likely originate from stem cells in the gastric cardia (19, 24), which over time appear to expand proximally into the squamous esophagus, leading eventually to dysplasia. The trigger for the activation and expansion of gastric cardia progenitors may be cytokine/chemokine signals from infiltrating immune cells. We noted increasing numbers of CXCR4⁺ epithelial cells at the SCJ in the CXCR4-GFP reporter mouse crossed to the pL2-IL1β mice (CXCR4-GFP;pL2-IL1β) during the stepwise progression from Barrett's esophagus to LGD and HGD (Fig. 1C) and mouse GFP⁺ epithelial cells (Fig. 2B). Many of these CXCR4⁻GFP⁺ cells were proliferating (Ki67⁺; Supplementary Fig. S2), and their numbers correlated with increasing grades of dysplasia. A similar increase in CXCR4⁺ epithelial columnar cells was observed in patients with Barrett's esophagus, with strong IHC staining particularly evident in Barrett's tissue with HGD (Fig. 1C). Thus, we further investigated the correlation of epithelial progenitor cells and elevated CXCR4 expression in our mouse model using FACS. Increasing numbers of CXCR4⁺ EpCAM⁺ cells were detected in the cardia and forestomach of mice with HGD lesions, compared with mice with LGD (Fig. 2H). In tumor microarrays from 59 patients with esophageal adenocarcinoma, CXCR4 expression was restricted to infiltrating immune cells in 53% of patients, in 3% of patients, CXCR4 was expressed only by epithelial tumor cells and in 42% of patients, both, immune cells and epithelial tumor cells contributed to the total CXCR4 expression (Fig. 2K and L). In summary, epithelial progenitor cells might represent yet another contributor of elevated CXCR4 expression in later stages of dysplasia, while immune cells are likely the main contributor to elevated CXCR4 expression in early dysplasia development.

The cellular proliferation and differentiation of epithelial cells depends on a large array of signaling molecules such as canonical Wnt (Wingless-Type MMTV Integration Site Family) signaling (25). As it is not clear whether such activation of Wnt signaling is critical in Barrett's metaplasia, dysplastic, and esophageal adenocarcinoma development, but WNT3 is suggested to play a key activation role in breast, rectal, lung, and gastric cancers through activation of the WNT–β-catenin–TCF canonical signaling pathway (26), we further investigated the Wnt signaling components in our Barrett's esophagus mouse model. In contrast to our findings in the gastric corpus, where Wnt5a-secreting innate lymphoid cells (ILC2) were critical to the development of diffuse gastric cancer (9), we found that in correlation to the increasing numbers of CXCR4 immune cells, Wnt3a but not Wnt5a was unregulated in the esophageal niche during accelerated tumorigenesis in our Barrett's esophagus model (Fig. 2J), pointing to Wnt3a as a potentially important cancer niche factor in the esophagus.

On the basis of the finding that local CXCR4 expression increased gradually during esophageal carcinogenesis, we hypothesized that CXCR4 expression mighty be exploited as a biomarker of disease progression. Thus, we conjugated a commercially available anti-mCXCR4 antibody with the NIR fluorescent dye Cy5.5 (AF680) and injected it intravenously into mice. The specificity of the fluorescent probe was validated in these studies by coinjecting the CXCR4-Cy5.5 antibody with the CXCR4 inhibitor AMD3100 (2 mg/kg), which blocked binding by the antibody (data not shown). Ex vivo fluorescence imaging at 24 hours after injection indeed showed that the CXCR4-Cy5.5 antibody specifically accumulated in small dysplastic lesions in pL2-IL1β mice, and that uptake was significantly increased in mice with HGD (high-fat diet pL2-IL1β; Figs. 3A, i and ii and 3B). Some CXCR4-specific Cy5.5 fluorescence signal was also detected at the SCJ (Fig. 3A, iii and iv). While there was a moderate increase of signal throughout the inflamed esophagus of pL2-IL1β mice, consistent with this receptor marking chronic inflammation, there was a remarkable high level of accumulation in areas of tumorigenesis (yellow lesions, marked with arrows), indicating that high levels of CXCR4 expression were very specific for the dysplastic lesion itself, reflecting more than simply an accumulation of chronic inflammatory cells.

[⁶⁸Ga]pentixafor is currently being evaluated clinically for PET imaging of CXCR4 expression in patients with cancer (LIT; refs. 27–29), and displays pronounced selectivity for human CXCR4. It does not bind to murine CXCR4, and thus is not suitable for in vivo imaging of disease progression in the pL2-IL1β transgenic mouse model. Therefore, an alternative CXCR4 ligand with high affinity for the murine CXCR4 receptor, [¹²⁵I]TUM-4007, a peptidic probe with a molecular weight <2,000 Da, was used to detect CXCR4 expression in murine Barrett's esophagus and esophageal adenocarcinoma via autoradiography. Mice were injected with approximately 1.85 MBq [¹²⁵I]TUM-4007 with or without 2 mg/kg AMD3100, and animals were sacrificed 1 hour later for biodistribution analysis. The ¹²⁵I-labeled mCXCR4-targeted ligand showed significant accumulation in the liver, along with moderate uptake in the spleen and stomach. A decrease in tracer accumulation in liver, spleen, and stomach was observed in mice coinjected with 2 mg/kg of the antagonist, AMD3100 (Fig. 4C), demonstrating CXCR4 specificity of radioligand uptake in these tissues. For ex vivo autoradiography, mice were injected with [¹²⁵I]TUM-4007 with or without AMD3100, sacrificed 1 hour postinjection, and CXCR4-mediated tracer accumulation in tissue cryosections from stomach and esophagus was visualized using autoradiography. Notably, high focal [¹²⁵I]TUM-4007 uptake was seen in pL2-IL1β mice with dysplasia, with a significant increase in tracer uptake in HGD (Fig. 4A). More importantly, a significant decrease in tracer signal was found in mice coinjected with 2 mg/kg of the inhibitor, AMD3100 (Fig. 4A and B). These data strongly suggest that uptake of the CXCR4 tracer is specific for tumor lesions and/or premalignant inflamed tissue, and correlates strongly with a dysplastic phenotype (Fig. 4A and B).

On the basis of our Barrett's esophagus mouse model data, we hypothesized that imaging of CXCR4 expression could be exploited as a marker for diagnosis and staging of esophageal adenocarcinoma in patients. As noted above, ex vivo fluorescence imaging in pL2-IL1β mice showed that CXCR4⁺ cells accumulated early on in the dysplastic lesion itself. The recent development of the CXCR4-targeted PET tracer [⁶⁸Ga]pentixafor (28, 29) has provided for sensitive and high-contrast imaging of hCXCR4 expression in patients suffering from lymphoma (30), multiple myeloma (27, 31), SCLC (32), glioblastoma (33), and other solid tumors (34, 35). Therefore, we investigated the potential of [⁶⁸Ga]pentixafor PET/CT for diagnosis and therapeutic monitoring of metastatic esophageal adenocarcinoma in a small patient cohort, and compared the findings with results using the metabolic tracer [¹⁸F]FDG, which is routinely used in the staging of esophageal adenocarcinoma.

[⁶⁸Ga]pentixafor-PET/CT was performed in five patients (one woman, four men; mean age, 72 ± 12.1 years) with esophageal cancer, and the studies confirmed that CXCR4 expression can be detected in vivo in esophageal adenocarcinoma (Fig. 5A). Overall, nine PET/CT examinations were carried out. [⁶⁸Ga]pentixafor was used in six and [¹⁸F]FDG in three scans. In three patients, both [⁶⁸Ga]pentixafor- and [¹⁸F]FDG-PET/CT imaging were performed for intrapatient comparison. One patient received [⁶⁸Ga]pentixafor-PET/CT imaging before and after chemotherapy. Another patient only received [⁶⁸Ga]pentixafor-PET/CT. Three of these patients showed metastasis. The evaluation of the SUV values is shown in Table 1. Representative patient data demonstrated strong signal intensity for both [18F]FDG-PET (n = 3) and [68Ga]pentixafor-PET (n = 5), but the latter appeared to show greater specificity; thus, while FDG-PET had a strong signal in the tumor region (Fig. 5, solid arrows), it also showed intense signals in the liver. In addition, CXCR4-mediated tracer uptake in the spleen was observed in patients with tumors (Fig. 5B, open arrow), confirming the observations in our mouse studies. Moreover, [⁶⁸Ga]pentixafor-PET/CT images of a patient with lymph node metastases indicated that CXCR4-targeted PET imaging might potentially be used to monitor metastasis in vivo (Fig. 5B). Furthermore, after receiving chemotherapy, the CXCR4 signal in both primary tumor and lymph node metastasis decreased (Fig. 5B). Finally, CXCR4 IHC of an esophageal adenocarcinoma from a patient imaged by [68Ga]CXCR4-PET/CT showed infiltration with CXCR4-expressing immune and epithelial cells (Supplementary Fig. S3), confirming that we indeed could image CXCR4 in patients with esophageal adenocarcinoma.

In this study, we show that local CXCR4 expression increases during the progression of dysplasia in a mouse model of Barrett's esophagus and esophageal adenocarcinoma. Mice harboring HGD demonstrated increased CXCR4 expression in preneoplastic tissue, compared with LGD. Interestingly, the increase in CXCR4 expression was observed largely in immune cells (neutrophils and T cells), but not fibroblast or endothelial cells, consistent with earlier observations that neutrophils are a key contributor to esophageal carcinogenesis (19). Increased numbers of neutrophils have been observed in patients with myxofibrosarcoma, gastric carcinoma, and melanoma (36–38). Neutrophils have been suggested to contribute to tumor cell growth and metastasis (39, 40), and increased neutrophil abundance predicts worse metastasis-specific survival in patients with breast cancer (41, 42). In esophageal cancer, a high neutrophil–lymphocyte ratio predicts a worse disease-free overall survival in surgically treated patients with esophageal squamous cell carcinoma (43), and in REAL-2–treated patients with advanced esophagogastric cancer (44).

It is likely that the CXCR4⁺ immune cells are recruited in part from the spleen, where CXCR4-mediated uptake of CXCR4-targeted imaging probes was observed both in humans and in our mouse model. Extramedullary contributions to the tumor microenvironment have been previously demonstrated, with studies showing that neutrophil precursors can physically relocate from the spleen to the tumor stroma (45). The stromal microenvironment of tumors includes a mixture of hematopoietic and mesenchymal cells, and clinical evidence supports the notion that these cells contribute to the development of a cancer (46). Specific targeting of such an environment using dedicated imaging probes may be used in the assessment of early tumor development, given that the recruitment of CXCR4⁺ neutrophils appears strongly associated with progression to HGD.

Furthermore, more CXCR4-expressing proliferating epithelial cells were observed in HGD mice, suggesting that epithelial progenitor cells also contribute to elevated CXCR4 expression in the later stages of dysplasia. In the mouse, CXCR4 is expressed in the proliferative zone of the cardia glands, such that the expansion in Barrett's esophagus again points to a cardia origin for this metaplastic lesion. As the numbers of infiltrating CXCR4⁺ immune cells were elevated in the early stages of dysplasia, the epithelial progenitors in the gastric cardia may possibly be activated by signals from the CXCR4⁺ immune cells or other stromal cells in the pretumorigenic microenvironment. Further studies, however, will be needed to define the source of the chemokine ligand, CXCL12 or SDF1, which presumably regulates the migration and expansion of CXCR4⁺ cells in Barrett's esophagus lesions.

The contribution of the recruited immune cells to the microenvironment, their interaction with progenitor epithelial cells, and their overall importance for tumor development or progression remain unresolved. Hayakawa and colleagues recently reported that a CXCL12/CXCR4 perivascular niche in diffuse-type gastric carcinogenesis supports normal and neoplastic stem cells through Wnt5a production (9). Given that Barrett's esophagus/esophageal adenocarcinoma derives from expansion of gastric cardia progenitors (19, 47), the role of the stromal niche seems highly relevant. Hayakawa and colleagues also demonstrated that a population of innate lymphoid cells (ILCs) expresses CXCR4 and that these are specifically recruited to the site of diffuse gastric cancer (DGC) arising in the corpus, and that depletion of such cells can inhibit tumorigenesis (9). Given the close relationship between esophageal and gastric cancer (48), we investigated the possible similarity between CXCR4⁺ immune cells in Barrett's esophagus and esophageal adenocarcinoma. However, in contrast to diffuse gastric cancer, the majority of CXCR4⁺ immune cells were not ILC2s but CD11b⁺ myeloid cells, and instead of an upregulation of Wnt5a as was observed in DGC, we found that Wnt3a was increased in the esophageal in our Barrett's esophagus mouse model. In addition, we recently showed that Wnt3a improved Barrett's organoid growth in 3D organoid cultures (49), underlining the likely contribution of Wnt3a to columnar epithelial expansion in the Barrett's esophagus mouse model. Wnt3a might well represent an important niche factor in Barrett's esophagus/esophageal adenocarcinoma, which would fit with its intestinal phenotype (50). Recent findings from our group have indicated that myeloid cells likely represent an important source of Wnt regulating proliferation in the gastrointestinal tract (51). In the canonical Wnt signaling pathway, Wnt proteins bind to the Frizzled/LRP receptor complex at the cell surface and further inhibit the degradation and subsequently accumulation of β-catenin in the cytoplasm and nucleus (52). It has been suggested that noncanonical Wnt signaling might inhibit Wnt/β-catenin signaling in developmental contexts. For example, Wnt5a expression can inhibit the constitutively high Wnt/β-catenin signaling activity of SW48 colon cancer cells (53). WNT5a may also act as tumor suppressor (54). Taken together, in our model, recruitment of stem cells and potential differentiation into intestinal metaplasia may rely more on myeloid cells and Wnt3a expression, similar to our current understanding for intestinal stem cells.

CXCR4-mediated probe accumulation was detected in both murine Barrett's esophagus and esophageal tumors by ex vivo fluorescence and tissue autoradiography. Moreover, in PET/CT imaging in human patients, the CXCR4-targeted PET probe [⁶⁸Ga]pentixafor accumulated in esophageal cancer tissue and in lymph node metastatic sites in esophageal adenocarcinoma patients. These data suggest that a distinct CXCR4⁺ immune or epithelial cell population might accumulate in tumors, although we have not been able yet to demonstrate a major fraction of tumor cells to be CXCR4⁺. Uptake of the CXCR4-targeted PET probe in malignant lesions was markedly reduced following chemotherapy, allowing us to hypothesize that advanced disease might be potentially monitored in this manner during treatment, similar to [¹⁸F]FDG-PET monitoring of treatment response (55), although confirmation of these preliminary data by more detailed studies is needed. Taken together, the initial results obtained from esophageal adenocarcinoma patients using [⁶⁸Ga]pentixafor-PET imaging are promising, and while comparable with standard [¹⁸F]FDG-PET with respect to imaging contrast and sensitivity of tumor detection, [⁶⁸Ga]pentixafor-PET provides complementary biochemical information. Moreover, in contrast to [¹⁸F]FDG, which accumulates in proliferating tumor tissue, [⁶⁸Ga]pentixafor-PET provides information on the tumor niche, and therefore might offer novel opportunities for cancer prevention and detection of early metastasis, although confirmation of these preliminary observations is needed. A much larger series of case studies will be required to confirm the ability to detect metastatic sites, and to determine whether PET-CT imaging can in fact detect early cancers or even dysplastic lesions. Given that in the mouse model, CXCR4⁺ immune cells are recruited during the process of carcinogenesis, we can speculate that such a cell type may help define the pretumor and/or premetastatic niche that precedes the local homing and proliferation of tumor cells. Given that Barrett's dysplasia likely involves an expansion of gastric cardia progenitors into the esophagus, the CXCR4⁺ immune cells may be crucial for the survival and expansion of migrating epithelial progenitors. In conclusion, CXCR4 (over)expression may characterize both tumor cells in esophageal tumors as well as cells in the local tumor microenvironment, making this receptor a promising and relatively specific molecular target for the detection, staging, and monitoring of esophageal adenocarcinoma in all stages of the disease.

H. Wester is a consultant/advisory board member for and holds ownership interest (including patents) in Scintomics GmbH. No potential conflicts of interest were disclosed by the other authors.

Conception and design: H.-Y. Fang, N.S. Münch, T.C. Wang, M. Quante

Development of methodology: H.-Y. Fang, N.S. Münch, M. Schottelius, J. Ingermann, H. Liu, S. Stangl, G. Multhoff, A.R. Sepulveda, H.-J. Wester

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.-Y. Fang, N.S. Münch, M. Schottelius, J. Ingermann, H. Liu, M. Schauer, S. Stangl, G. Multhoff, K. Steiger, A.A. Kühl, M. Quante

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.-Y. Fang, N.S. Münch, M. Schottelius, M. Schauer, S. Stangl, K. Steiger, C. Gerngroß, M. Jesinghaus, M. Quante

Writing, review, and/or revision of the manuscript: H.-Y. Fang, N.S. Münch, M. Schottelius, S. Stangl, G. Multhoff, C. Gerngroß, W. Weichert, A.A. Kühl, A.R. Sepulveda, H.-J. Wester, T.C. Wang, M. Quante

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N.S. Münch, K. Steiger, C. Gerngroß, H.-J. Wester

Study supervision: H.-Y. Fang, N.S. Münch, H.-J. Wester, M. Quante

The research leading to these results has received funding from the Deutsche Forschungsgemeinschaft (DFG) under grant agreement no. SFB 824, has been funded by the Deutsche Krebshilfe with in the Max Eder Program (to M. Quante), and 2 U54 CA163004-06 (principal investigator; The role of the microenvironment in Barrett's esophagus, within the BETRNet program).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.